Merge remote-tracking branch 'origin/main' into r6955

2026-07-26 12:52:57 +00:00 · 2026-07-26 12:52:57 +00:00 · d9efdafaa7
commit d9efdafaa7
parent ffa28dc34b 0c1c9f71db
624 changed files with 67183 additions and 11902 deletions
--- a/.github/workflows/cross-platform-parity-ci.yml
+++ b/.github/workflows/cross-platform-parity-ci.yml
@ -1,18 +1,16 @@
 # SPDX-License-Identifier: AGPL-3.0-only
 # Copyright 2026-present the Unsloth AI Inc. team. All rights reserved.

-# Runs installer parity and autostart opt-out tests on Windows and macOS.
+# Runs installer parity and autostart opt-out tests across all three platforms.
 #
-# Why: that test is the guard that install.sh and install.ps1 stay in
-# sync, but today it only runs on ubuntu-latest (auto-discovered by
-# studio-backend-ci.yml's "Repo tests (CPU)" job). The test reads both
-# installer scripts, and on Windows Path.read_text() defaults to the
-# cp1252 locale encoding, so a non-cp1252 byte in install.sh (it already
-# contains a U+274C) raises UnicodeDecodeError there even though Linux and
-# macOS default to UTF-8. The reads were pinned to encoding="utf-8" in
-# #6166; this job keeps that from silently regressing by exercising the
-# test on the platforms it claims parity for. Pure pytest, no GPU,
-# sub-second, so the matrix is cheap.
+# Why: the parity test guards that install.sh and install.ps1 stay in sync.
+# It originally ran only on ubuntu-latest through studio-backend-ci.yml.
+# On Windows, Path.read_text() defaults to the cp1252 locale encoding, so a
+# non-cp1252 byte in install.sh raises UnicodeDecodeError even though Linux
+# and macOS default to UTF-8. The reads were pinned to encoding="utf-8" in
+# #6166; this matrix keeps that from silently regressing. Pure pytest, no GPU,
+# sub-second, so the matrix is cheap. Linux also runs the POSIX rollback test
+# under dash, matching the supported curl-to-sh installer path.

 name: Cross-platform parity

@ -23,6 +21,8 @@ on:
      - 'install.ps1'
      - 'tests/test_installer_skip_autostart.py'
      - 'tests/python/test_cross_platform_parity.py'
+      - 'tests/sh/test_install_rollback_lifecycle.sh'
+      - 'tests/studio/test_install_rollback_lifecycle.ps1'
      - '.github/workflows/cross-platform-parity-ci.yml'
  push:
    branches: [main]
@ -31,6 +31,8 @@ on:
      - 'install.ps1'
      - 'tests/test_installer_skip_autostart.py'
      - 'tests/python/test_cross_platform_parity.py'
+      - 'tests/sh/test_install_rollback_lifecycle.sh'
+      - 'tests/studio/test_install_rollback_lifecycle.ps1'
      - '.github/workflows/cross-platform-parity-ci.yml'
  workflow_dispatch:

@ -47,7 +49,7 @@ jobs:
    strategy:
      fail-fast: false
      matrix:
-        os: [windows-latest, macos-latest]
+        os: [ubuntu-latest, windows-latest, macos-latest]
    runs-on: ${{ matrix.os }}
    timeout-minutes: 10
    steps:
@ -67,3 +69,10 @@ jobs:
          tests/python/test_cross_platform_parity.py
          tests/test_installer_skip_autostart.py
          -q
+      - name: PowerShell rollback lifecycle tests
+        if: runner.os == 'Windows'
+        shell: pwsh
+        run: pwsh -NoProfile -File tests/studio/test_install_rollback_lifecycle.ps1
+      - name: POSIX rollback lifecycle tests
+        if: runner.os == 'Linux'
+        run: sh tests/sh/test_install_rollback_lifecycle.sh
--- a/.github/workflows/studio-backend-ci.yml
+++ b/.github/workflows/studio-backend-ci.yml
@ -30,6 +30,13 @@ on:
      - 'unsloth/**'
      - 'unsloth_cli/**'
      - 'tests/**'
+      # The root installers: tests/sh/*.sh and tests/studio/install/* assert
+      # against these two files, so a change here must run the suite that
+      # covers it. Without them an install-only edit (the shape most AMD/ROCm
+      # routing fixes take) skipped Backend CI entirely.
+      - 'install.sh'
+      - 'install.ps1'
+      - 'scripts/**'
      - 'pyproject.toml'
      - '.github/workflows/studio-backend-ci.yml'
  push:
@ -193,6 +200,7 @@ jobs:
            --ignore=tests/sh \
            --ignore=tests/studio/test_hardware_dispatch_matrix.py \
            --ignore=tests/studio/test_is_mlx_dispatch_gate.py \
+            --ignore=tests/studio/test_xpu_spoof_pipeline.py \
            --ignore=tests/vllm_compat \
            --ignore=tests/version_compat \
            -m 'not server and not e2e' \
@ -205,36 +213,43 @@ jobs:
        env:
          PYTHONPATH: ${{ github.workspace }}/studio
          UNSLOTH_COMPILE_DISABLE: '1'
-        # These two files mutate hardware.py module globals at runtime
-        # via the spoof fixtures, which leaks state into any other test
-        # that imports hardware. Run them in their own pytest invocation
-        # so the leak does not cross file boundaries.
+        # These files mutate hardware.py module globals at runtime via the
+        # spoof fixtures (CUDA/ROCm/XPU/MLX/CPU), which leaks state into any
+        # other test that imports hardware. Run them in their own pytest
+        # invocation so the leak does not cross file boundaries.
        run: |
          python -m pytest -q --tb=short \
            tests/studio/test_hardware_dispatch_matrix.py \
-            tests/studio/test_is_mlx_dispatch_gate.py
+            tests/studio/test_is_mlx_dispatch_gate.py \
+            tests/studio/test_xpu_spoof_pipeline.py

      - name: Shell installer tests
-        # Subset that does not depend on a writable / pristine install.sh
-        # tree; test_install_host_defaults.sh checks install.ps1 layout
-        # which has drifted (separate followup).
+        # Auto-discovered rather than allowlisted. The old hardcoded list had
+        # silently fallen seven files behind tests/run_all.sh, including
+        # test_strixhalo_wsl_reroute.sh -- the only shell coverage of the ROCm
+        # WSL reroute -- so that suite never ran on a PR. Skips are explicit,
+        # each with a reason, and tests/studio/test_ci_shell_suite_coverage.py
+        # fails if this step stops discovering the directory or the skip list
+        # grows without one.
+        #
+        # Skipped:
+        #   test_install_host_defaults.sh: asserts an install.ps1 layout that
+        #     has drifted (separate followup).
+        #   test_install_rollback_lifecycle.sh: already runs on both platforms
+        #     in cross-platform-parity-ci.yml.
        run: |
          set -e
-          for s in \
-              tests/sh/test_get_torch_index_url.sh \
-              tests/sh/test_mac_intel_compat.sh \
-              tests/sh/test_node_decision.sh \
-              tests/sh/test_studio_home_node_dir.sh \
-              tests/sh/test_system_node_readonly.sh \
-              tests/sh/test_nvcc_meets_llama_minimum.sh \
-              tests/sh/test_resolve_cuda_archs.sh \
-              tests/sh/test_tauri_install_exit_order.sh \
-              tests/sh/test_torch_constraint.sh \
-              tests/sh/test_torch_flavor.sh \
-              tests/sh/test_with_llama_cpp_dir_flag.sh \
-              tests/sh/test_with_llama_cpp_dir_link_behavior.sh; do
+          skip="test_install_host_defaults.sh test_install_rollback_lifecycle.sh"
+          found=0
+          for s in tests/sh/test_*.sh; do
+              case " $skip " in
+                  *" $(basename "$s") "*) echo "skipping $s (see workflow comment)"; continue ;;
+              esac
+              found=$((found + 1))
              echo "::group::$s"
              bash "$s"
              echo "::endgroup::"
          done
+          [ "$found" -gt 0 ] || { echo "::error::no shell tests discovered under tests/sh"; exit 1; }
+          echo "ran $found shell installer test files"

--- a/.github/workflows/studio-ui-smoke.yml
+++ b/.github/workflows/studio-ui-smoke.yml
@ -231,12 +231,69 @@ jobs:
          mkdir -p logs/playwright_extra
          python tests/studio/playwright_extra_ui.py

+      - name: UI font size scaling regression (Playwright)
+        env:
+          BASE_URL: http://127.0.0.1:18894
+          STUDIO_PW: ${{ env.STUDIO_EXTRA_NEW_PW }}
+          PW_ART_DIR: logs/playwright_fontscale
+        run: |
+          mkdir -p logs/playwright_fontscale
+          python tests/studio/playwright_ui_font_scale.py
+
      - name: Stop second Unsloth
        if: always()
        run: |
          kill "${STUDIO_EXTRA_PID}" 2>/dev/null || true
          sleep 2

+      # Model-picker per-model-config regression (PR #7207 re-land of #6647).
+      # Fourth Unsloth on its own port; loads the tiny GGUF and drives the
+      # picker's run-settings surface: Context Length persists across a reload,
+      # Reset clears the stored override (never pins it), and the infra models
+      # (RAG embedder + llama.cpp probe) stay hidden from the picker.
+      - name: Reset auth + boot Unsloth for model-config tests (port 18898)
+        run: |
+          unsloth studio reset-password
+          mkdir -p logs
+          UNSLOTH_API_ONLY=1 unsloth studio -H 127.0.0.1 -p 18898 \
+            > logs/studio_modelcfg.log 2>&1 &
+          echo "STUDIO_MODELCFG_PID=$!" >> "$GITHUB_ENV"
+
+      - name: Wait for /api/health on 18898
+        run: |
+          for i in $(seq 1 180); do
+            if curl -fs "http://127.0.0.1:18898/api/health" > /tmp/health4.json; then
+              jq -e '.status == "healthy"' /tmp/health4.json && break
+            fi
+            sleep 1
+          done
+          jq -e '.status == "healthy"' /tmp/health4.json
+
+      - name: Pass bootstrap pw for model-config test
+        run: |
+          NEW="CIModelCfg-$(python -c 'import secrets; print(secrets.token_urlsafe(16))')"
+          echo "::add-mask::$NEW"
+          echo "STUDIO_MODELCFG_NEW_PW=$NEW" >> "$GITHUB_ENV"
+
+      - name: Drive model-picker per-model-config with Playwright
+        env:
+          BASE_URL: http://127.0.0.1:18898
+          STUDIO_NEW_PW: ${{ env.STUDIO_MODELCFG_NEW_PW }}
+          PW_ART_DIR: logs/playwright_modelcfg
+          STUDIO_UI_STRICT: '1'
+          GGUF_REPO: ${{ env.GGUF_REPO }}
+          GGUF_VARIANT: ${{ env.GGUF_VARIANT }}
+          STUDIO_MODEL_HINT: gemma-3-270m
+        run: |
+          mkdir -p logs/playwright_modelcfg
+          python tests/studio/playwright_model_config.py
+
+      - name: Stop fourth Unsloth
+        if: always()
+        run: |
+          kill "${STUDIO_MODELCFG_PID}" 2>/dev/null || true
+          sleep 2
+
      # IME + multilingual paste regression (issue #5318 / PR #5327).
      # Third Unsloth on its own port so a hang here cannot poison the
      # earlier UI tests. No GGUF -- the bug surface is the composer.
@ -297,12 +354,15 @@ jobs:
          path: |
            logs/studio.log
            logs/studio_extra.log
+            logs/studio_modelcfg.log
            logs/studio_ime.log
            logs/install.log
            logs/server-logs/
            logs/playwright
            logs/playwright-permissions-*
            logs/playwright_extra
+            logs/playwright_fontscale
+            logs/playwright_modelcfg
            logs/playwright_ime
            logs/studio-permissions-*.log
          retention-days: 7
--- a/README.md
+++ b/README.md
@ -86,6 +86,13 @@ Replace `claude` with any supported agent:
 | OpenCode | `unsloth start opencode` |
 | Pi Coding Agent | `unsloth start pi` |

+Claude Code, Codex, OpenCode and Pi can keep their current model and use Unsloth as a local
+subagent:
+
+```bash
+unsloth start claude --as-subagent --model unsloth/model-GGUF:quant
+```
+
 ## 📥 Install
 Unsloth can be used in two ways: through **[Unsloth Studio](https://unsloth.ai/docs/new/studio/)**, the web UI, or through **Unsloth Core**, the code-based version. Each has different requirements.

--- a/install.ps1
+++ b/install.ps1
@ -1416,13 +1416,82 @@ exit 0
            $suffix++
            $candidate = Join-Path $StudioHome "unsloth_studio.rollback.$stamp.$PID.$suffix"
        }
-        Move-Item -LiteralPath $ExistingDir -Destination $candidate -ErrorAction Stop
        $script:StudioVenvRollbackDir = $candidate
        $script:StudioVenvRollbackTarget = $ExistingDir
        $script:StudioVenvRollbackActive = $true
+        # Publish the rollback state before the atomic rename so interruption
+        # cannot land after Move-Item but before cleanup knows where the old venv went.
+        try {
+            Move-Item -LiteralPath $ExistingDir -Destination $candidate -ErrorAction Stop
+        } catch {
+            # A collision or ordinary rename failure leaves the original in place.
+            # Keep state active only when the rename happened before interruption.
+            if (Test-Path -LiteralPath $ExistingDir) {
+                $script:StudioVenvRollbackActive = $false
+                $script:StudioVenvRollbackDir = $null
+            }
+            throw
+        }
        substep "previous environment preserved for rollback"
    }

+    function Remove-StudioVenvTreeWithRetry {
+        param(
+            [Parameter(Mandatory = $true)][string]$Path,
+            [Parameter(Mandatory = $true)][string]$Label
+        )
+        $lastError = $null
+        for ($attempt = 1; $attempt -le 3; $attempt++) {
+            try {
+                Remove-Item -LiteralPath $Path -Recurse -Force -ErrorAction Stop
+            } catch {
+                $lastError = $_.Exception.Message
+            }
+            if (-not (Test-Path -LiteralPath $Path)) { return $true }
+            if ($attempt -lt 3) { Start-Sleep -Milliseconds (250 * $attempt) }
+        }
+        Write-Host "[WARN] Could not remove $Label at $Path" -ForegroundColor Yellow
+        if ($lastError) { Write-Host "       $lastError" -ForegroundColor Yellow }
+        return $false
+    }
+
+    function Test-StudioVenvRollbackMustBePreserved {
+        param([Parameter(Mandatory = $true)][System.IO.FileSystemInfo]$Rollback)
+        # Preserve anything outside the installer's timestamp.PID[.suffix] format.
+        if ($Rollback.Name -notmatch '^unsloth_studio\.rollback\.[0-9]{14}\.([0-9]+)(?:\.[0-9]+)?$') {
+            return $true
+        }
+        $ownerPid = 0
+        if (-not [int]::TryParse($Matches[1], [ref]$ownerPid)) { return $true }
+        if ($ownerPid -eq $PID) { return $true }
+        return $null -ne (Get-Process -Id $ownerPid -ErrorAction SilentlyContinue)
+    }
+
+    function Remove-StaleStudioVenvRollbacks {
+        try {
+            $rollbacks = @(
+                Get-ChildItem -LiteralPath $StudioHome -Directory -Force -ErrorAction Stop |
+                    Where-Object { $_.Name -like 'unsloth_studio.rollback.*' }
+            )
+        } catch {
+            Write-Host "[WARN] Could not inspect stale environment rollbacks in $StudioHome" -ForegroundColor Yellow
+            Write-Host "       $($_.Exception.Message)" -ForegroundColor Yellow
+            return
+        }
+        foreach ($rollback in $rollbacks) {
+            if (($rollback.Attributes -band [System.IO.FileAttributes]::ReparsePoint) -ne 0) {
+                Write-Host "[WARN] Refusing to remove rollback reparse point $($rollback.FullName)" -ForegroundColor Yellow
+                continue
+            }
+            # A concurrent installer may have moved its live venv aside. The PID
+            # in the generated name keeps this run from deleting its rescue copy.
+            if (Test-StudioVenvRollbackMustBePreserved -Rollback $rollback) { continue }
+            if (Remove-StudioVenvTreeWithRetry -Path $rollback.FullName -Label "stale environment rollback") {
+                substep "removed stale environment rollback $($rollback.Name)"
+            }
+        }
+    }
+
    function Restore-StudioVenvRollback {
        if (-not $script:StudioVenvRollbackActive) { return }
        $backup = $script:StudioVenvRollbackDir
@ -1434,7 +1503,9 @@ exit 0
        substep "restoring previous environment after failed install..." "Yellow"
        try {
            if (Test-Path -LiteralPath $target) {
-                Remove-Item -LiteralPath $target -Recurse -Force -ErrorAction SilentlyContinue
+                if (-not (Remove-StudioVenvTreeWithRetry -Path $target -Label "incomplete environment")) {
+                    throw "Could not remove incomplete environment at $target"
+                }
            }
            Move-Item -LiteralPath $backup -Destination $target -Force -ErrorAction Stop
            substep "restored previous environment"
@ -1449,13 +1520,17 @@ exit 0
    function Complete-StudioVenvRollback {
        if (-not $script:StudioVenvRollbackActive) { return }
        $backup = $script:StudioVenvRollbackDir
-        if ($backup -and (Test-Path -LiteralPath $backup)) {
-            Remove-Item -LiteralPath $backup -Recurse -Force -ErrorAction SilentlyContinue
-        }
+        # The replacement is committed. Disable restoration before deleting the
+        # backup so interruption cannot restore a partially deleted environment.
        $script:StudioVenvRollbackActive = $false
        $script:StudioVenvRollbackDir = $null
+        if ($backup -and (Test-Path -LiteralPath $backup)) {
+            Remove-StudioVenvTreeWithRetry -Path $backup -Label "environment rollback" | Out-Null
+        }
    }

+    $studioVenvReplacementCommitted = $false
+    try {
    if (Test-Path -LiteralPath $VenvPython) {
        # why: matching guard to the .venv branch below -- in env-mode
        # $StudioHome is a user-chosen workspace, so refuse to nuke an
@ -1842,12 +1917,14 @@ exit 0
            #    (gfx120X/110X/1151/1150/103X); unknown names fall back to CPU.
            elseif ($ROCmGpuLabel) {
                $nameArchTable = @(
-                    @{ P = "9070 XT|9080";                                        A = "gfx1201" }  # RDNA 4 (RX 9070 XT / 9080)
-                    @{ P = "9070|9060";                                           A = "gfx1200" }  # RDNA 4 (RX 9070 / 9060)
-                    @{ P = "8060S|8050S|8040S|Strix Halo|Ryzen AI Max|AI Max"; A = "gfx1151" }  # RDNA 3.5 (Strix Halo: Radeon 8060S/8050S/8040S iGPU, Ryzen AI Max+)
-                    @{ P = "890M|880M|860M|840M|Strix Point|Krackan|HX 37[05]|AI 9 HX|AI 9 36[05]|AI 7 35[05]|AI 5 34[05]|AI 7 PRO 35|AI 5 33"; A = "gfx1150" }  # RDNA 3.5 (Strix/Krackan Point: Radeon 890M/880M iGPU, Ryzen AI 9 HX 370/375)
-                    @{ P = "RX 7900|RX 7800|RX 7700(?!S)|PRO W7900|PRO W7800|PRO W7700"; A = "gfx1100" }  # RDNA 3 desktop/workstation (Navi 31)
-                    @{ P = "RX 7600|RX 7700S|RX 7650|PRO W7600|PRO W7500|PRO V710"; A = "gfx1102" }  # RDNA 3 (Navi 33)
+                    @{ P = "9070|9080";                                           A = "gfx1201" }  # RDNA 4 (Navi 48: RX 9070 XT / 9070 GRE / 9070 / 9080)
+                    @{ P = "9060";                                                A = "gfx1200" }  # RDNA 4 (Navi 44: RX 9060 XT / 9060)
+                    @{ P = "8065S|8060S|8050S|8040S|Strix Halo|Ryzen AI Max|AI Max"; A = "gfx1151" }  # RDNA 3.5 (Strix Halo + Gorgon Halo: Radeon 8065S/8060S/8050S/8040S iGPU, Ryzen AI Max / Max+)
+                    @{ P = "890M|880M|Strix Point|HX 37[05]|AI 9 HX|AI 9 36[05]"; A = "gfx1150" }  # RDNA 3.5 (Strix Point: Radeon 890M/880M, Ryzen AI 9 HX 370/375)
+                    @{ P = "860M|840M|Krackan|AI 7 35[05]|AI 5 34[05]|AI 7 PRO 35|AI 5 33"; A = "gfx1152" }  # RDNA 3.5 (Krackan Point: Radeon 860M/840M, Ryzen AI 7 350 / AI 5 340)
+                    @{ P = "RX 7900|PRO W7900|PRO W7800";                         A = "gfx1100" }  # RDNA 3 desktop/workstation (Navi 31)
+                    @{ P = "RX 7800|RX 7700(?!S)|PRO W7700|PRO V710";             A = "gfx1101" }  # RDNA 3 (Navi 32)
+                    @{ P = "RX 7600|RX 7700S|RX 7650|PRO W7600|PRO W7500";        A = "gfx1102" }  # RDNA 3 (Navi 33)
                    @{ P = "780M|760M|740M|Phoenix|Hawk Point|Z1 Extreme|Z2 Extreme"; A = "gfx1103" }  # RDNA 3 iGPU (Phoenix / Hawk Point)
                    @{ P = "RX 6900|RX 6800|RX 6750|RX 6700|PRO W6800|PRO W6900";  A = "gfx1030" }  # RDNA 2 (Navi 21) -- gfx103X family
                    @{ P = "RX 6650|RX 6600|PRO W6600|PRO W6650";                  A = "gfx1032" }  # RDNA 2 (Navi 23) -- gfx103X family
@ -2030,16 +2107,18 @@ exit 0
    # _strip_index_url_credentials (install.sh / py / setup.ps1).
    function Remove-IndexUrlCredentials {
        param([string]$Url)
-        $sep = $Url.IndexOf('://')
+        # Ordinal, not culture-aware: on non-English locales (e.g. th-TH) linguistic
+        # IndexOf treats "://" as ignorable, mis-locates it, and crashes Substring (issue #7279).
+        $sep = $Url.IndexOf('://', [System.StringComparison]::Ordinal)
        if ($sep -lt 0) { return $Url }
        $scheme = $Url.Substring(0, $sep)
        $rest = $Url.Substring($sep + 3)
        # Drop query / fragment (may hold auth tokens).
        $q = $rest.IndexOfAny([char[]]('?', '#'))
        if ($q -ge 0) { $rest = $rest.Substring(0, $q) }
-        $slash = $rest.IndexOf('/')
+        $slash = $rest.IndexOf('/', [System.StringComparison]::Ordinal)
        $authority = if ($slash -ge 0) { $rest.Substring(0, $slash) } else { $rest }
-        $at = $authority.LastIndexOf('@')
+        $at = $authority.LastIndexOf('@', [System.StringComparison]::Ordinal)
        $host_ = if ($at -ge 0) { $authority.Substring($at + 1) } else { $authority }
        if ($slash -ge 0) { return "${scheme}://${host_}$($rest.Substring($slash))" }
        return "${scheme}://${host_}"
@ -2126,8 +2205,13 @@ exit 0
        $archFamilyMap = @{
            "gfx1201" = "gfx120X-all"; "gfx1200" = "gfx120X-all"  # RDNA 4
            "gfx1151" = "gfx1151";     "gfx1150" = "gfx1150"       # RDNA 3.5 (Strix Halo/Point)
+            "gfx1152" = "gfx1152"                                  # RDNA 3.5 (Krackan Point)
            "gfx1103" = "gfx110X-all"; "gfx1102" = "gfx110X-all"   # RDNA 3
            "gfx1101" = "gfx110X-all"; "gfx1100" = "gfx110X-all"
+            "gfx1036" = "gfx103X-all"; "gfx1035" = "gfx103X-all"   # RDNA 2 (RX 6000)
+            "gfx1034" = "gfx103X-all"; "gfx1033" = "gfx103X-all"
+            "gfx1032" = "gfx103X-all"; "gfx1031" = "gfx103X-all"
+            "gfx1030" = "gfx103X-all"
            "gfx90a"  = "gfx90a";      "gfx908"  = "gfx908"        # MI200/MI100
        }
        # gfx120X (RDNA 4) and gfx1151/gfx1150 (Strix) have a null-pointer bug in
@ -2143,6 +2227,7 @@ exit 0
        $torchFloorMap = @{
            "gfx1201" = "torch>=2.11.0,<2.12.0"; "gfx1200" = "torch>=2.11.0,<2.12.0"
            "gfx1151" = "torch>=2.11.0,<2.12.0"; "gfx1150" = "torch>=2.11.0,<2.12.0"
+            "gfx1152" = "torch>=2.11.0,<2.12.0"
        }
        # Companion ranges track the torch ceiling so pip resolves a consistent
        # trio on AMD's per-arch index (each published independently). Mirrors
@ -2150,10 +2235,12 @@ exit 0
        $torchvisionFloorMap = @{
            "gfx1201" = "torchvision>=0.26.0,<0.27.0"; "gfx1200" = "torchvision>=0.26.0,<0.27.0"
            "gfx1151" = "torchvision>=0.26.0,<0.27.0"; "gfx1150" = "torchvision>=0.26.0,<0.27.0"
+            "gfx1152" = "torchvision>=0.26.0,<0.27.0"
        }
        $torchaudioFloorMap = @{
            "gfx1201" = "torchaudio>=2.11.0,<2.12.0"; "gfx1200" = "torchaudio>=2.11.0,<2.12.0"
            "gfx1151" = "torchaudio>=2.11.0,<2.12.0"; "gfx1150" = "torchaudio>=2.11.0,<2.12.0"
+            "gfx1152" = "torchaudio>=2.11.0,<2.12.0"
        }
        $archFamily = if ($ROCmGfxArch -and $archFamilyMap.ContainsKey($ROCmGfxArch)) { $archFamilyMap[$ROCmGfxArch] } else { $null }
        if ($archFamily) {
@ -2183,7 +2270,7 @@ exit 0
            $_pinRocm211 = ([int]$Matches[1] -eq 7 -and [int]$Matches[2] -eq 2)
        }
        # Only the 2.11-allowlist gfx arches need the floor; others publish <2.11 and stay bare.
-        $_pinGfx211 = @('gfx120x-all', 'gfx1151', 'gfx1150') -contains $_pinLeaf
+        $_pinGfx211 = @('gfx120x-all', 'gfx1151', 'gfx1150', 'gfx1152') -contains $_pinLeaf
        if ($_pinGfx211 -or $_pinRocm211) {
            $ROCmIndexUrl = $TorchIndexUrl
            $ROCmTorchFloor = "torch>=2.11.0,<2.12.0"
@ -2266,7 +2353,7 @@ exit 0
        if ($SkipTorch) {
            # No-torch: install unsloth + unsloth-zoo with --no-deps, then
            # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (migrated no-torch)" { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" }
+            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (migrated no-torch)" { uv pip install --python $VenvPython --no-deps --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
            if ($baseInstallExit -eq 0) {
                # Resolve pydantic WITH deps so pip pins pydantic-core
                # to the matching version (no-torch-runtime.txt below
@ -2280,7 +2367,7 @@ exit 0
                }
            }
        } else {
-            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (migrated)" { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" }
+            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (migrated)" { uv pip install --python $VenvPython --reinstall-package unsloth --reinstall-package unsloth-zoo "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
        }
        if ($baseInstallExit -ne 0) {
            Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
@ -2354,7 +2441,7 @@ exit 0
        if ($SkipTorch) {
            # No-torch: install unsloth + unsloth-zoo with --no-deps, then
            # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
-            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (no-torch)" { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" }
+            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (no-torch)" { uv pip install --python $VenvPython --no-deps --upgrade-package unsloth --upgrade-package unsloth-zoo "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
            if ($baseInstallExit -eq 0) {
                # Same pydantic-with-deps trick as the migrated branch.
                $baseInstallExit = Invoke-InstallCommandRetry -Label "install pydantic" { uv pip install --python $VenvPython pydantic }
@ -2366,7 +2453,7 @@ exit 0
                }
            }
        } elseif ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (local)" { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" }
+            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (local)" { uv pip install --python $VenvPython --upgrade-package unsloth "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
        } else {
            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth" { uv pip install --python $VenvPython --upgrade-package unsloth -- "$PackageName" }
        }
@ -2394,7 +2481,7 @@ exit 0
        Write-TauriLog "STEP" "Installing unsloth"
        substep "installing unsloth (this may take a few minutes)..."
        if ($StudioLocalInstall) {
-            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (auto torch backend)" { uv pip install --python $VenvPython "unsloth-zoo>=2026.7.4" "unsloth>=2026.7.4" --torch-backend=auto }
+            $baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (auto torch backend)" { uv pip install --python $VenvPython "unsloth-zoo>=2026.7.6" "unsloth>=2026.7.5" --torch-backend=auto }
            if ($baseInstallExit -ne 0) {
                Write-Host "[ERROR] Failed to install unsloth (exit code $baseInstallExit)" -ForegroundColor Red
                return (Exit-InstallFailure "Failed to install unsloth (exit code $baseInstallExit)" $baseInstallExit)
@ -2420,6 +2507,13 @@ exit 0
        }
    }

+    $installedPackageVersion = (& $VenvPython -c "from importlib.metadata import version; import sys; print(version(sys.argv[1]))" $PackageName 2>$null | Out-String).Trim()
+    if ($LASTEXITCODE -eq 0 -and $installedPackageVersion) {
+        step $PackageName "$installedPackageVersion installed"
+    } else {
+        substep "[WARN] installed $PackageName version could not be determined" "Yellow"
+    }
+
    # ── Enforce the installed torch flavor matches the detected GPU build ──
    # PEP 440 ignores the +cpu/+cuXXX/+rocm local label in a version range, so uv
    # keeps a stale torch==X+cpu against a CUDA index and setup.ps1 then loops on
@ -2675,6 +2769,13 @@ exit 0
    }
    Refresh-SessionPath  # sync current session with registry
    Complete-StudioVenvRollback
+    $studioVenvReplacementCommitted = $true
+    Remove-StaleStudioVenvRollbacks
+    } finally {
+        if (-not $studioVenvReplacementCommitted) {
+            Restore-StudioVenvRollback
+        }
+    }

    # Env-mode session export AFTER Refresh-SessionPath; otherwise a legacy
    # User PATH entry (Machine > User > current $env:Path) would win.
--- a/install.sh
+++ b/install.sh
@ -475,14 +475,20 @@ _start_studio_venv_replacement() {
    _stamp=$(date +%Y%m%d%H%M%S 2>/dev/null || echo "time")
    _candidate="$STUDIO_HOME/unsloth_studio.rollback.$_stamp.$$"
    _suffix=0
-    while [ -e "$_candidate" ]; do
+    while [ -e "$_candidate" ] || [ -L "$_candidate" ]; do
        _suffix=$((_suffix + 1))
        _candidate="$STUDIO_HOME/unsloth_studio.rollback.$_stamp.$$.$_suffix"
    done
-    mv "$_existing_dir" "$_candidate"
    _VENV_ROLLBACK_DIR="$_candidate"
    _VENV_ROLLBACK_TARGET="$_existing_dir"
    _VENV_ROLLBACK_ACTIVE=true
+    # Publish the rollback state before the atomic rename so a signal cannot
+    # land after mv but before the exit handlers know where the old venv went.
+    if ! mv "$_existing_dir" "$_candidate"; then
+        _VENV_ROLLBACK_ACTIVE=false
+        _VENV_ROLLBACK_DIR=""
+        return 1
+    fi
    substep "previous environment preserved for rollback"
 }

@ -503,13 +509,68 @@ _restore_studio_venv_replacement() {
    fi
 }

-_commit_studio_venv_replacement() {
-    [ "$_VENV_ROLLBACK_ACTIVE" = true ] || return 0
-    if [ -n "$_VENV_ROLLBACK_DIR" ] && [ -d "$_VENV_ROLLBACK_DIR" ]; then
-        rm -rf "$_VENV_ROLLBACK_DIR" || true
+_studio_venv_rollback_must_be_preserved() {
+    _rollback_name=${1##*/}
+    _rollback_metadata=${_rollback_name#unsloth_studio.rollback.}
+    _rollback_stamp=${_rollback_metadata%%.*}
+    _rollback_process=${_rollback_metadata#*.}
+    # Preserve anything outside the installer's timestamp.PID[.suffix] format.
+    [ "$_rollback_process" != "$_rollback_metadata" ] || return 0
+    case "$_rollback_stamp" in
+        time) ;;
+        ''|*[!0-9]*) return 0 ;;
+        *) [ "${#_rollback_stamp}" -eq 14 ] || return 0 ;;
+    esac
+    _rollback_pid=${_rollback_process%%.*}
+    case "$_rollback_pid" in
+        ''|*[!0-9]*) return 0 ;;
+    esac
+    _rollback_suffix=${_rollback_process#*.}
+    if [ "$_rollback_suffix" != "$_rollback_process" ]; then
+        case "$_rollback_suffix" in ''|*[!0-9]*) return 0 ;; esac
    fi
-    _VENV_ROLLBACK_ACTIVE=false
-    _VENV_ROLLBACK_DIR=""
+    kill -0 "$_rollback_pid" 2>/dev/null
+}
+
+_prune_stale_studio_venv_rollbacks() {
+    for _stale_rollback in "$STUDIO_HOME"/unsloth_studio.rollback.*; do
+        [ -d "$_stale_rollback" ] || continue
+        if [ -L "$_stale_rollback" ]; then
+            echo "⚠️  Refusing to remove rollback symlink $_stale_rollback" >&2
+            continue
+        fi
+        # A concurrent installer may have moved its live venv aside. The PID in
+        # the generated name keeps this successful run from deleting its rescue copy.
+        _studio_venv_rollback_must_be_preserved "$_stale_rollback" && continue
+        if rm -rf "$_stale_rollback"; then
+            substep "removed stale environment rollback ${_stale_rollback##*/}"
+        else
+            echo "⚠️  Could not remove stale environment rollback $_stale_rollback" >&2
+        fi
+    done
+}
+
+_commit_studio_venv_replacement() {
+    if [ "$_VENV_ROLLBACK_ACTIVE" = true ]; then
+        _rollback_to_remove="$_VENV_ROLLBACK_DIR"
+        # The new environment is already committed. Clear the restore state
+        # before deletion so an interrupt cannot replace it with a half-deleted backup.
+        _VENV_ROLLBACK_ACTIVE=false
+        _VENV_ROLLBACK_DIR=""
+        if [ -n "$_rollback_to_remove" ] && [ -d "$_rollback_to_remove" ]; then
+            if ! rm -rf "$_rollback_to_remove"; then
+                echo "⚠️  Could not remove environment rollback $_rollback_to_remove" >&2
+            fi
+        fi
+    fi
+    # Only prune older orphaned copies after the replacement has succeeded, so
+    # an interrupted install never discards the last known-good environment.
+    _prune_stale_studio_venv_rollbacks
+}
+
+_cleanup_install_temporaries() {
+    [ -n "${_UV_OVERRIDE_TMPDIR:-}" ] && rm -rf "$_UV_OVERRIDE_TMPDIR" 2>/dev/null || true
+    [ -n "${_UNSLOTH_TORCH_OVERRIDES:-}" ] && rm -f "$_UNSLOTH_TORCH_OVERRIDES" 2>/dev/null || true
 }

 _on_install_exit() {
@ -517,15 +578,28 @@ _on_install_exit() {
    if [ "$_status" -ne 0 ]; then
        _restore_studio_venv_replacement
    fi
-    [ -n "${_UV_OVERRIDE_TMPDIR:-}" ] && rm -rf "$_UV_OVERRIDE_TMPDIR" 2>/dev/null || true
-    [ -n "${_UNSLOTH_TORCH_OVERRIDES:-}" ] && rm -f "$_UNSLOTH_TORCH_OVERRIDES" 2>/dev/null || true
+    _cleanup_install_temporaries
    exit "$_status"
 }
+
+_on_install_signal() {
+    _signal_status="$1"
+    # EXIT is disabled to avoid a second cleanup pass. Ignore further termination
+    # signals until the old environment is back in place.
+    trap - EXIT
+    trap '' HUP INT TERM
+    _restore_studio_venv_replacement
+    _cleanup_install_temporaries
+    exit "$_signal_status"
+}
 # Empty so an inherited value never reaches the trap's rm; only temp paths this
 # script creates below (spaced-path dir, torch-trio overrides) are removed.
 _UV_OVERRIDE_TMPDIR=""
 _UNSLOTH_TORCH_OVERRIDES=""
 trap _on_install_exit EXIT
+trap '_on_install_signal 129' HUP
+trap '_on_install_signal 130' INT
+trap '_on_install_signal 143' TERM

 # ── Helper: download a URL to a file (supports curl and wget) ──
 download() {
@ -551,6 +625,45 @@ _is_pkg_installed() {
    esac
 }

+# ── Helper: human-readable apt distro label for the sudo package prompt (#6207) ──
+# Reads /etc/os-release so the Accept? prompt can say which distro we detected and
+# that packages come from that distro's official apt repos (not a tarball).
+_apt_distro_description() {
+    # Plain ( ... ) subshell — not $() — so case/;; stays bash-3.2-safe on macOS.
+    # Bash 3.2 misparses case arms inside command substitution and errors on `;;`.
+    (
+        if [ ! -r /etc/os-release ]; then
+            printf 'a debian-like system'
+            exit 0
+        fi
+        # shellcheck disable=SC1091
+        . /etc/os-release 2>/dev/null || true
+        if [ -n "${NAME:-}" ] && [ -n "${VERSION_ID:-}" ]; then
+            _ad_label="$NAME $VERSION_ID"
+        elif [ -n "${PRETTY_NAME:-}" ]; then
+            _ad_label="$PRETTY_NAME"
+        elif [ -n "${NAME:-}" ]; then
+            _ad_label="$NAME"
+        else
+            printf 'a debian-like system'
+            exit 0
+        fi
+        case " ${ID:-} ${ID_LIKE:-} " in
+            *" debian "*|*" ubuntu "*) _ad_label="${_ad_label} (debian-like)" ;;
+        esac
+        printf '%s' "$_ad_label"
+    )
+}
+
+# ── Helper: can the controlling terminal actually be opened for reading? ──
+# `test -r` only checks permission bits, which look fine in containers and
+# systemd units where open() then fails with ENXIO. Probe with a real open.
+# The subshell is required: in dash a failed redirection on the special
+# builtin `:` exits the whole script.
+_can_read_tty() {
+    ( : </dev/tty ) >/dev/null 2>&1
+}
+
 # ── Helper: install packages via apt, escalating to sudo only if needed ──
 # Usage: _smart_apt_install pkg1 pkg2 pkg3 ...
 _smart_apt_install() {
@ -581,31 +694,73 @@ _smart_apt_install() {

    # Step 3: Escalate -- need elevated permissions for remaining packages
    if command -v sudo >/dev/null 2>&1; then
+        _ad_desc="$(_apt_distro_description)"
        echo ""
        echo "    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
        echo "    WARNING: We require sudo elevated permissions to install:"
        echo "    $_STILL_MISSING"
-        echo "    If you accept, we'll run sudo now, and it'll prompt your password."
+        echo "    Detected ${_ad_desc}."
+        echo "    If you accept, we'll run sudo apt-get to install these packages"
+        echo "    from your distro's official repositories (not a third-party tarball)."
        echo "    !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
        echo ""
-        printf "    Accept? [Y/n] "
-        if [ -r /dev/tty ]; then
-            read -r REPLY </dev/tty || REPLY="y"
-        else
-            REPLY="y"
-        fi
-        case "$REPLY" in
-            [nN]*)
+        if _can_read_tty; then
+            printf "    Accept? [Y/n] "
+            # The device opened, so a failed read is EOF, not consent: decline,
+            # as the autostart prompt below does. Enter is still yes (a
+            # successful read of an empty line).
+            read -r REPLY </dev/tty || REPLY="n"
+            case "$REPLY" in
+                [nN]*)
+                    echo ""
+                    echo "    Please install these packages first, then re-run Unsloth Studio setup:"
+                    echo "    sudo apt-get update -y && sudo apt-get install -y $_STILL_MISSING"
+                    exit 1
+                    ;;
+            esac
+            # Mirror the headless branch: on a sudoers denial, a wrong password
+            # or an apt error, say what to run by hand instead of letting set -e
+            # abort on a bare sudo/apt message.
+            if sudo apt-get update -y </dev/null &&
+                sudo apt-get install -y $_STILL_MISSING </dev/null; then
+                :
+            else
                echo ""
-                echo "    Please install these packages first, then re-run Unsloth Studio setup:"
+                echo "    Could not install these packages: $_STILL_MISSING"
+                echo "    See the error above."
+                echo "    Please install them first, then re-run Unsloth Studio setup:"
                echo "    sudo apt-get update -y && sudo apt-get install -y $_STILL_MISSING"
                exit 1
-                ;;
-            *)
-                sudo apt-get update -y </dev/null
-                sudo apt-get install -y $_STILL_MISSING </dev/null
-                ;;
-        esac
+            fi
+        else
+            # Nobody can answer a prompt or type a password here. -n makes sudo
+            # refuse rather than prompt into a closed stdin, which is how #7307
+            # died. Probe with the real commands: `sudo -l` answers whether they
+            # are *authorized*, not whether running them needs authentication.
+            # -k ignores any cached timestamp, so only a real NOPASSWD rule gets
+            # through, not someone's sudo in another shell minutes ago. Per
+            # sudo(8), -k alongside a command ignores the cached credentials and
+            # "will not update" them, so other sessions keep theirs.
+            echo "    No terminal to confirm on; trying passwordless sudo."
+            if sudo -n -k apt-get update -y </dev/null &&
+                sudo -n -k apt-get install -y $_STILL_MISSING </dev/null; then
+                echo "    Installed with passwordless sudo."
+            else
+                echo ""
+                echo "    Could not install these packages: $_STILL_MISSING"
+                echo "    Detected ${_ad_desc}."
+                # Either sudo refused, or apt failed on a bad repo, dpkg lock or
+                # network outage. sudo exits 1 on an auth/config problem and
+                # when the command cannot be executed, but otherwise passes the
+                # command's own status through, so state both causes.
+                echo "    Either sudo needs a password here, or apt-get itself"
+                echo "    failed; see the error above. With no terminal to"
+                echo "    authenticate on, this cannot be done unattended."
+                echo "    Please install them first, then re-run Unsloth Studio setup:"
+                echo "    sudo apt-get update -y && sudo apt-get install -y $_STILL_MISSING"
+                exit 1
+            fi
+        fi
    else
        echo ""
        echo "    sudo is not available on this system."
@ -1636,7 +1791,7 @@ _maybe_reroute_strixhalo_to_2404() {
    # CUDA_VISIBLE_DEVICES=""/-1 and the /proc/driver/nvidia fallback for PATH/timeout gaps.
    if _has_usable_nvidia_gpu; then return 0; fi
    # Strix APUs show in /proc/cpuinfo; discrete cards don't, so also try WMI. Either reroutes.
-    if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9]0S|Strix Halo' /proc/cpuinfo 2>/dev/null \
+    if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9][05]S|Strix Halo' /proc/cpuinfo 2>/dev/null \
       && ! _wsl_amd_gpu_name >/dev/null 2>&1; then
        return 0
    fi
@ -2115,18 +2270,155 @@ _has_amd_rocm_gpu() {
         amd-smi list 2>/dev/null | awk '/^GPU[[:space:]]*[:\[][[:space:]]*[0-9]/{ found=1 } END{ exit !found }'; then
        return 0
    elif [ -e /dev/kfd ] && \
-         awk 'FNR==1{ gpu=0; amd=0 } /gpu_id/{ gpu=($2+0>0) } /vendor_id/{ amd=($2==4098) } \
-              gpu && amd { found=1 } END{ exit !found }' \
+         awk '/vendor_id/ && $2 == 4098 { found = 1 } END { exit !found }' \
             /sys/class/kfd/kfd/topology/nodes/*/properties 2>/dev/null; then
-        # vendor_id 4098 = 0x1002 (AMD). NVIDIA open kernel module (driver
-        # 560+) can register KFD topology nodes with non-zero gpu_id but
-        # vendor_id 4318 (0x10DE). Require AMD vendor to avoid misrouting
-        # NVIDIA-only hosts to the ROCm install path.
+        # vendor_id 4098 = 0x1002 (AMD) marks a GPU node: the KFD CPU node
+        # reports vendor_id 0, so any 4098 node is an AMD GPU. NVIDIA's open
+        # kernel module (driver 560+) registers KFD nodes as vendor_id 4318
+        # (0x10DE), so this never false-positives on NVIDIA-only hosts.
+        # The prior check also required a gpu_id line, but gpu_id is a SIBLING
+        # sysfs file, not a line in properties -- it never matched, so the
+        # fallback silently missed every ROCm-less AMD host (issue: fresh
+        # Arch/CachyOS boxes reporting "no GPU detected").
        return 0
    fi
    return 1
 }

+# Returns 0 if an AMD display GPU is on the PCI bus even when ROCm can't use it
+# (e.g. a Strix Halo iGPU with no /dev/kfd). Only sharpens the "no GPU detected"
+# hint. vendor 0x1002 = AMD/ATI; class 0x03* = display controller.
+_amd_gpu_present_via_pci() {
+    [ -d /sys/bus/pci/devices ] || return 1
+    for _pci_vendor in /sys/bus/pci/devices/*/vendor; do
+        [ -r "$_pci_vendor" ] || continue
+        read -r _v < "$_pci_vendor" 2>/dev/null || continue
+        [ "$_v" = "0x1002" ] || continue
+        _cls="${_pci_vendor%vendor}class"
+        [ -r "$_cls" ] || continue
+        read -r _c < "$_cls" 2>/dev/null || continue
+        case "$_c" in 0x03*) return 0 ;; esac
+    done
+    return 1
+}
+
+# Map a gfx arch to the AMD pip index family (mirrors install.ps1 $archFamilyMap).
+_amd_arch_index_family_for_gfx() {
+    case "$1" in
+        gfx1201|gfx1200) echo gfx120X-all ;;
+        gfx1151) echo gfx1151 ;;
+        gfx1150) echo gfx1150 ;;
+        gfx1152) echo gfx1152 ;;
+        gfx1103|gfx1102|gfx1101|gfx1100) echo gfx110X-all ;;
+        gfx1036|gfx1035|gfx1034|gfx1033|gfx1032|gfx1031|gfx1030) echo gfx103X-all ;;
+        gfx90a) echo gfx90a ;;
+        gfx908) echo gfx908 ;;
+        *) return 1 ;;
+    esac
+}
+
+# Map a GPU marketing name to gfx arch (kept in sync with install.ps1 nameArchTable).
+_infer_amd_gfx_arch_from_gpu_name() {
+    case "$1" in
+        *9070*|*9080*) echo gfx1201 ;;
+        *9060*) echo gfx1200 ;;
+        *"8065S"*|*"8060S"*|*"8050S"*|*"8040S"*|*"Strix Halo"*|*"Ryzen AI Max"*|*"AI Max"*) echo gfx1151 ;;
+        *"890M"*|*"880M"*|*"Strix Point"*|*"HX 37"*|*"AI 9 HX"*|*"AI 9 36"*) echo gfx1150 ;;
+        *"860M"*|*"840M"*|*"Krackan"*|*"AI 7 35"*|*"AI 5 34"*|*"AI 7 PRO 35"*|*"AI 5 33"*) echo gfx1152 ;;
+        *"RX 7600"*|*"RX 7700S"*|*"RX 7650"*|*"PRO W7600"*|*"PRO W7500"*) echo gfx1102 ;;
+        *"RX 7800"*|*"RX 7700"*|*"PRO W7700"*|*"PRO V710"*) echo gfx1101 ;;
+        *"RX 7900"*|*"PRO W7900"*|*"PRO W7800"*) echo gfx1100 ;;
+        *"780M"*|*"760M"*|*"740M"*|*"Phoenix"*|*"Hawk Point"*|*"Z1 Extreme"*|*"Z2 Extreme"*) echo gfx1103 ;;
+        *"RX 6900"*|*"RX 6800"*|*"RX 6750"*|*"RX 6700"*|*"PRO W6800"*|*"PRO W6900"*) echo gfx1030 ;;
+        *"RX 6650"*|*"RX 6600"*|*"PRO W6600"*|*"PRO W6650"*) echo gfx1032 ;;
+        *"RX 6500"*|*"RX 6400"*|*"RX 6300"*|*"PRO W6400"*|*"PRO W6500"*) echo gfx1034 ;;
+        *) return 1 ;;
+    esac
+}
+
+# Best-effort gfx inference when ROCm tools can't see the GPU (unslothai#7301).
+# Mirrors install.ps1 arch resolution on Windows ($HasROCm false, $ROCmGfxArch set).
+_infer_linux_amd_gfx_arch() {
+    if [ -n "${UNSLOTH_ROCM_GFX_ARCH:-}" ]; then
+        printf '%s\n' "$(printf '%s' "$UNSLOTH_ROCM_GFX_ARCH" | tr '[:upper:]' '[:lower:]')"
+        return 0
+    fi
+    # On WSL /proc/cpuinfo and lspci still report the host APU, but without the
+    # ROCDXG bridge (librocdxg over /dev/dxg) the AMD wheels can't reach the GPU;
+    # keep the CPU fallback there unless that runtime is present (the explicit
+    # override above still wins). Mirrors install_python_stack.py.
+    _gpu_evidence=""
+    if [ -e /dev/dxg ] || grep -qi microsoft /proc/version 2>/dev/null; then
+        for _d in /opt/rocm/lib /opt/rocm/lib64 /opt/rocm-*/lib /opt/rocm-*/lib64; do
+            { [ -e "$_d/librocdxg.so" ] || [ -e "$_d/librocdxg.so.1" ]; } && _rocdxg=1 && break
+        done
+        [ -n "${_rocdxg:-}" ] || return 1
+        # WSL enumerates no PCI display device; /dev/dxg + librocdxg IS the
+        # GPU evidence there.
+        _gpu_evidence=1
+    elif _amd_gpu_present_via_pci; then
+        _gpu_evidence=1
+    fi
+    # /proc/cpuinfo leaks the HOST CPU model into VMs/containers that received
+    # no AMD GPU, so the CPU-model text alone is not GPU evidence: require an
+    # AMD display device (PCI vendor 0x1002, class 0x03*) before trusting it.
+    # The lspci fallback below needs no gate; an AMD display line IS evidence.
+    if [ -n "$_gpu_evidence" ] && grep -qiE 'Ryzen AI Max|Radeon 80[0-9][05]S|Strix Halo' /proc/cpuinfo 2>/dev/null; then
+        echo gfx1151
+        return 0
+    fi
+    if [ -n "$_gpu_evidence" ] && grep -qiE '890M|880M|Strix Point|HX 37[05]|AI 9 HX|AI 9 36[05]' /proc/cpuinfo 2>/dev/null; then
+        echo gfx1150
+        return 0
+    fi
+    if [ -n "$_gpu_evidence" ] && grep -qiE '860M|840M|Krackan|AI 7 35[05]|AI 5 34[05]|AI 7 PRO 35|AI 5 33' /proc/cpuinfo 2>/dev/null; then
+        echo gfx1152
+        return 0
+    fi
+    if command -v lspci >/dev/null 2>&1; then
+        # A non-AMD controller can enumerate first (Intel/ASPEED before an AMD
+        # dGPU), so scan every display-class line and take the first AMD one
+        # that maps. The vendor guard is case-SENSITIVE (a -i "ATI" would match
+        # "CorporATIon" on every Intel/NVIDIA line); whole-line matching also
+        # survives the 0000: PCI domain prefix. Mirrors install_python_stack.py.
+        _amd_disp=$(lspci -nn 2>/dev/null | grep -E 'VGA compatible controller|3D controller|Display controller' | grep -E 'AMD|ATI' || true)
+        while IFS= read -r _ln; do
+            [ -n "$_ln" ] || continue
+            if _gfx=$(_infer_amd_gfx_arch_from_gpu_name "$_ln"); then
+                echo "$_gfx"
+                return 0
+            fi
+        done <<EOF
+$_amd_disp
+EOF
+    fi
+    return 1
+}
+
+# Reads the AMD gfx arch for wheel-index decisions: a user-set
+# UNSLOTH_ROCM_GFX_ARCH is authoritative (lowercased), else rocminfo, then
+# amd-smi. rocminfo/amd-smi honor ROCR/HIP_VISIBLE_DEVICES, so a container mask
+# (e.g. ROCR_VISIBLE_DEVICES=-1) would hide a GPU that the env-independent KFD
+# detection still sees -- the tool probes run with the masks cleared. Prints the
+# gfx token(s) or nothing when unreadable, and always returns 0 (a failing probe
+# as the last command would trip set -e in callers' assignments). Shared by
+# get_torch_index_url's gfx gate and the runtime-less reroute gate so the two
+# can never disagree on what "readable" means.
+_probe_amd_gfx_arch() {
+    _ensure_rocm_probe_env
+    _pg=$(printf '%s' "${UNSLOTH_ROCM_GFX_ARCH:-}" | tr '[:upper:]' '[:lower:]')
+    if [ -z "$_pg" ] && command -v rocminfo >/dev/null 2>&1; then
+        _pg=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; rocminfo 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+    fi
+    if [ -z "$_pg" ] && command -v amd-smi >/dev/null 2>&1; then
+        _pg=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; amd-smi list 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+        if [ -z "$_pg" ]; then
+            _pg=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; amd-smi static --asic 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+        fi
+    fi
+    printf '%s\n' "$_pg"
+}
+
 # ── Detect GPU and choose PyTorch index URL ──
 # Mirrors Get-TorchIndexUrl in install.ps1.
 # On CPU-only machines this returns the cpu index, avoiding the solver
@ -2180,6 +2472,29 @@ get_torch_index_url() {
        if ! _has_amd_rocm_gpu; then
            echo "$_base/cpu"; return
        fi
+        # A generic rocm index is only safe when the gfx arch is readable: the
+        # Strix reroute (gfx1150/1151 -> arch-specific index) learns gfx from
+        # rocminfo/amd-smi, so if those are missing OR do not enumerate the GPU, an
+        # unknown-arch box might be Strix and would get the broken _grouped_mm
+        # wheels. Probe via the shared helper (override first, then rocminfo/amd-smi
+        # with visibility masks cleared); if the arch is unreadable, never guess a
+        # rocm index. A KFD-only host whose arch is still inferable from hardware
+        # IDs (PCI/cpuinfo/lspci) returns the cpu index and lets the runtime-less
+        # reroute below upgrade it to AMD per-arch wheels -- the reroute gate uses
+        # this same probe, so the handoff can't misfire. Only when inference fails
+        # too is CPU final, with the actionable warning.
+        _amd_gfx_probe=$(_probe_amd_gfx_arch)
+        if [ -z "$_amd_gfx_probe" ]; then
+            if _amd_inferred_gfx=$(_infer_linux_amd_gfx_arch 2>/dev/null) && \
+               [ -n "$_amd_inferred_gfx" ] && \
+               _amd_arch_index_family_for_gfx "$_amd_inferred_gfx" >/dev/null 2>&1; then
+                echo "[WARN] AMD GPU detected but rocminfo/amd-smi can't read its gfx arch -- inferring $_amd_inferred_gfx from hardware IDs." >&2
+                echo "$_base/cpu"; return
+            fi
+            echo "[WARN] AMD GPU detected but its gfx arch can't be read (rocminfo/amd-smi missing or not enumerating the GPU) -- installing CPU-only PyTorch." >&2
+            echo "[WARN] For GPU PyTorch, install or repair rocminfo/amd-smi (e.g. sudo pacman -S rocm-hip-sdk) and re-run this installer." >&2
+            echo "$_base/cpu"; return
+        fi
        # AMD GPU confirmed -- detect ROCm version
        _rocm_tag=""
        _rocm_tag=$({ command -v amd-smi >/dev/null 2>&1 && \
@ -2196,7 +2511,11 @@ get_torch_index_url() {
            { command -v rpm >/dev/null 2>&1 && \
                ver="$(rpm -q --qf '%{VERSION}\n' rocm-core 2>/dev/null)" && \
                [ -n "$ver" ] && \
-                printf '%s\n' "$ver" | awk -F'[.-]' '{print "rocm"$1"."$2; exit}'; }) 2>/dev/null
+                printf '%s\n' "$ver" | awk -F'[.-]' '{print "rocm"$1"."$2; exit}'; }) 2>/dev/null || _rocm_tag=""
+        # ^ || guard: when EVERY version source is missing (e.g. rocminfo present
+        # but rocm-core not installed, so dpkg-query/rpm exit 1), the whole ||
+        # chain fails and set -e would kill the installer BEFORE the actionable
+        # no-version WARN below -- exactly the fresh-install case it exists for.
        # Validate _rocm_tag: must match "rocmX.Y" with major >= 1
        case "$_rocm_tag" in
            rocm[1-9]*.[0-9]*) : ;;  # valid (major >= 1)
@ -2232,12 +2551,27 @@ get_torch_index_url() {
            esac
            return
        fi
-        # AMD GPU confirmed by rocminfo/amd-smi but ROCm version could not be
-        # read from any source (amd-smi, /opt/rocm/.info/version, hipconfig,
-        # dpkg, rpm).  Warn explicitly rather than silently installing CPU PyTorch.
-        echo "[WARN] AMD GPU detected but ROCm version could not be determined -- falling back to CPU-only PyTorch" >&2
-        echo "[WARN] Ensure one of the following is accessible: amd-smi, hipconfig, /opt/rocm/.info/version, rocm-core package" >&2
-        echo "[WARN] To install ROCm: https://rocm.docs.amd.com/en/latest/deploy/linux/index.html" >&2
+        # AMD GPU confirmed (rocminfo/amd-smi or the KFD topology fallback) but
+        # no ROCm/HIP install was found to read the version from (amd-smi,
+        # /opt/rocm/.info/version, hipconfig, dpkg, rpm). This is the common
+        # fresh-install case: the GPU is real, but with no ROCm userspace the
+        # correct PyTorch build can't be selected. Warn with an actionable fix
+        # rather than silently installing CPU PyTorch.
+        # A user-set UNSLOTH_ROCM_GFX_ARCH seeded the probe above, so rocminfo/
+        # amd-smi may still be unable to see the GPU; when the named arch maps to
+        # a wheel family, the runtime-less reroute (gated on the override) will
+        # install the AMD per-arch wheels -- a CPU-only warning here would be
+        # false for that path. Defer like the inferable-arch branch does.
+        if [ -n "${UNSLOTH_ROCM_GFX_ARCH:-}" ] && \
+           _amd_arch_index_family_for_gfx "$_amd_gfx_probe" >/dev/null 2>&1; then
+            echo "[WARN] AMD GPU detected with no readable ROCm version, but UNSLOTH_ROCM_GFX_ARCH=$_amd_gfx_probe is set -- routing to AMD per-arch wheels." >&2
+            echo "$_base/cpu"; return
+        fi
+        echo "[WARN] AMD GPU detected, but no ROCm/HIP install was found to select the matching GPU PyTorch build -- falling back to CPU-only PyTorch." >&2
+        echo "[WARN] Install the ROCm/HIP SDK, then re-run this installer:" >&2
+        echo "[WARN]   Arch / CachyOS : sudo pacman -S rocm-hip-sdk" >&2
+        echo "[WARN]   other distros  : https://rocm.docs.amd.com/en/latest/deploy/linux/index.html" >&2
+        echo "[WARN] Minimum required for version detection: amd-smi, hipconfig, /opt/rocm/.info/version, or the rocm-core package." >&2
        echo "$_base/cpu"; return
    fi
    # Parse CUDA version from nvidia-smi output (POSIX-safe, no grep -P).
@ -2649,7 +2983,7 @@ _maybe_bootstrap_rocm_wsl() {
    [ -e /dev/dxg ] || return 0
    # Strix APUs show in /proc/cpuinfo (the CPU model); discrete cards don't, so also
    # ask the Windows host. Either signal suffices; the bootstrap detects arch from rocminfo.
-    if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9]0S|Strix Halo' /proc/cpuinfo 2>/dev/null \
+    if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9][05]S|Strix Halo' /proc/cpuinfo 2>/dev/null \
       && ! _wsl_amd_gpu_name >/dev/null 2>&1; then
        return 0
    fi
@ -2735,6 +3069,72 @@ fi

 TORCH_INDEX_URL=$(get_torch_index_url)

+# Linux: ROCm runtime missing but a supported AMD gfx arch is inferable (Strix Halo
+# in /proc/cpuinfo, lspci marketing name, UNSLOTH_ROCM_GFX_ARCH). Route to AMD's
+# per-arch wheels like install.ps1 does on Windows (unslothai#7301).
+# Gated on the runtime probes NOT naming a gfx: either no AMD GPU is detected at
+# all (_has_amd_rocm_gpu false), or the GPU is visible only through the
+# env-independent KFD topology while rocminfo/amd-smi can't read its arch
+# (KFD-only host, unslothai#7314 -- before the KFD detection fix these hosts
+# reached this reroute via the false branch, so the empty-probe condition
+# preserves that routing). A */cpu index chosen WITH a readable gfx
+# (unsupported/unreadable ROCm version, after its own warning) is a deliberate
+# fallback -- rerouting it would contradict that decision, and stays excluded
+# because the shared probe returns its gfx. An explicit UNSLOTH_ROCM_GFX_ARCH
+# override stays authoritative either way.
+if [ "$_torch_index_pinned" = false ] && [ "$SKIP_TORCH" = false ] && \
+   ! _has_usable_nvidia_gpu && \
+   { [ -n "${UNSLOTH_ROCM_GFX_ARCH:-}" ] || ! _has_amd_rocm_gpu || \
+     [ -z "$(_probe_amd_gfx_arch)" ]; } && \
+   case "$(uname -s)" in Linux) true ;; *) false ;; esac && \
+   case "$_ARCH" in x86_64|amd64) true ;; *) false ;; esac; then
+    # ROCm torch wheels are x86_64-only; get_torch_index_url returns CPU on other
+    # arches, so an inferred/overridden gfx must not reroute arm64 to AMD wheels.
+    case "$TORCH_INDEX_URL" in
+        */cpu)
+            _linux_inferred_gfx=$(_infer_linux_amd_gfx_arch 2>/dev/null || true)
+            if [ -n "$_linux_inferred_gfx" ]; then
+                _amd_family=$(_amd_arch_index_family_for_gfx "$_linux_inferred_gfx") || _amd_family=""
+                if [ -n "$_amd_family" ]; then
+                    _amd_mirror="${UNSLOTH_AMD_ROCM_MIRROR:-https://repo.amd.com/rocm/whl}"
+                    while [ "${_amd_mirror%/}" != "$_amd_mirror" ]; do
+                        _amd_mirror="${_amd_mirror%/}"
+                    done
+                    TORCH_INDEX_URL="${_amd_mirror}/${_amd_family}/"
+                    # Hand the inferred arch to setup.sh (llama.cpp): it re-probes
+                    # ROCm on its own, and on these runtime-less hosts its probes
+                    # find nothing, so without this it classifies the box as
+                    # non-ROCm and installs the CPU prebuilt while torch just got
+                    # AMD per-arch wheels. setup.sh and install_llama_prebuilt.py
+                    # both honor UNSLOTH_ROCM_GFX_ARCH, so exporting it is the
+                    # whole handoff (a user-set override re-exports unchanged).
+                    export UNSLOTH_ROCM_GFX_ARCH="$_linux_inferred_gfx"
+                    case "$_linux_inferred_gfx" in
+                        gfx1201|gfx1200|gfx1151|gfx1150|gfx1152)
+                            TORCH_CONSTRAINT="torch>=2.11.0,<2.12.0"
+                            TORCHVISION_CONSTRAINT="torchvision>=0.26.0,<0.27.0"
+                            TORCHAUDIO_CONSTRAINT="torchaudio>=2.11.0,<2.12.0"
+                            ;;
+                    esac
+                    echo "" >&2
+                    # KFD-only hosts reach this reroute with /dev/kfd present
+                    # (that's what detected them), so don't claim it's missing.
+                    if _has_amd_rocm_gpu; then
+                        echo "  [WARN] AMD GPU visible via the kernel driver (KFD) but rocminfo/amd-smi can't read its gfx arch; using $_linux_inferred_gfx." >&2
+                    else
+                        echo "  [WARN] ROCm runtime not visible (/dev/kfd, rocminfo, amd-smi) but $_linux_inferred_gfx inferred." >&2
+                    fi
+                    echo "  [WARN] Routing to AMD arch-specific wheels ($(_strip_index_url_credentials "$TORCH_INDEX_URL"))." >&2
+                    echo "  [WARN] These wheels bundle their own ROCm runtime; install the kernel stack for native compute:" >&2
+                    echo "  [WARN]   https://docs.unsloth.ai/get-started/install-and-update/amd" >&2
+                    echo "  [WARN] Tip: set UNSLOTH_ROCM_GFX_ARCH=$_linux_inferred_gfx to skip inference next time." >&2
+                    echo "" >&2
+                fi
+            fi
+            ;;
+    esac
+fi
+
 # Export the resolved torch backend ("cuda", "rocm", or "cpu") so that
 # downstream scripts (setup.sh -> install_python_stack.py) know what was
 # chosen here and can skip ROCm-specific repair steps on CUDA/CPU hosts.
@ -2779,7 +3179,7 @@ fi
 # and a bare name can resolve a 2.12 ABI-mismatched wheel. Match on the FINAL leaf so a
 # custom mirror with a gfx/rocm7.2 path segment but a cu*/cpu family isn't forced.
 case "$_torch_index_leaf" in
-    rocm7.2|gfx120x-all|gfx1151|gfx1150)
+    rocm7.2|gfx120x-all|gfx1151|gfx1150|gfx1152)
        TORCH_CONSTRAINT="torch>=2.11.0,<2.12.0"
        TORCHVISION_CONSTRAINT="torchvision>=0.26.0,<0.27.0"
        TORCHAUDIO_CONSTRAINT="torchaudio>=2.11.0,<2.12.0"
@ -2818,29 +3218,64 @@ case "$TORCH_INDEX_URL" in
        fi
        ;;
 esac
-# ── Strix Halo / Strix Point: force rocm7.2 wheels, bypass Radeon repo ───────
-# gfx1151 (Strix Halo) and gfx1150 (Strix Point) have a ROCm 7.1 driver bug
-# that causes a segfault in torch._grouped_mm (moe_utils.py line 167).
-# The Radeon repo now ships cp313 wheels for rocm-rel-7.1, so when
-# _amd_gpu_radeon=true the installer silently lands on the broken combo.
-# Detect these GPUs when TORCH_INDEX_URL is rocm7.1 and override to rocm7.2.
-case "$TORCH_INDEX_URL" in
-    */rocm7.1|*/rocm7.1.*)
+# 0 when a rocmX.Y index leaf ($1, the final path segment) is older than floor
+# $2.$3 (int compare, so rocm7.2 < rocm7.13). Non-rocm leaves (gfx*, cu*, cpu) and
+# non-numeric versions return 1. Leaf-based (like $_torch_index_leaf) so a mirror
+# base holding its own rocm token compares the family leaf, not the base path.
+_rocm_leaf_below() {
+    case "$1" in rocm[0-9]*.[0-9]*) : ;; *) return 1 ;; esac
+    _rb=${1#rocm}; _maj=${_rb%%.*}; _min=${_rb#*.}; _min=${_min%%.*}
+    case "$_maj$_min" in *[!0-9]*) return 1 ;; esac
+    if [ "$_maj" -lt "$2" ]; then return 0; fi
+    if [ "$_maj" -eq "$2" ] && [ "$_min" -lt "$3" ]; then return 0; fi
+    return 1
+}
+# ── Strix Halo / Strix Point: route to the AMD arch-specific index ───────────
+# gfx1151/gfx1150 need torch 2.11+rocm7.13 from repo.amd.com/rocm/whl/gfx<arch>/,
+# which carries AMD's real fixes (the rocm7.1 _grouped_mm segfault, moe_utils.py:167,
+# and later Strix kernel bugs). Every generic pytorch.org index below rocm7.13 lacks
+# them (and the Radeon repo can be offline, unslothai#7264), so reroute a detected
+# Strix GPU whenever the picked index is older than the arch build -- covers today's
+# rocm6.0-7.2 and any future 7.x < 7.13; rocm7.13+ already has the fixes, so leave it.
+case "$_torch_index_leaf" in
+    rocm[0-9]*)
        # Collect every gfx token in rocminfo / amd-smi enumeration order
        # (skip duplicates), then index by HIP_VISIBLE_DEVICES /
        # ROCR_VISIBLE_DEVICES so a mixed Strix iGPU + non-Strix dGPU box
        # where the user selected the dGPU does NOT get rerouted to the
        # Strix per-gfx index.
-        _gfx_all=""
-        if command -v rocminfo >/dev/null 2>&1; then
-            _gfx_all=$(rocminfo 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}')
+        # || true on each probe: no gfx match makes grep exit 1, which under
+        # set -euo pipefail would abort the installer before the next fallback
+        # runs (now that the case matches every rocm* index, not just rocm7.1).
+        # A user-supplied UNSLOTH_ROCM_GFX_ARCH overrides probing (mirrors setup.sh
+        # and the display block), so a Strix override still reaches the arch index.
+        _gfx_all=$(printf '%s' "${UNSLOTH_ROCM_GFX_ARCH:-}" | tr '[:upper:]' '[:lower:]')
+        if [ -z "$_gfx_all" ] && command -v rocminfo >/dev/null 2>&1; then
+            _gfx_all=$(rocminfo 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
        fi
        if [ -z "$_gfx_all" ] && command -v amd-smi >/dev/null 2>&1; then
-            _gfx_all=$(amd-smi list 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}')
+            _gfx_all=$(amd-smi list 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
            # PowerShell paths also probe `amd-smi static --asic`; mirror it
            # so a host with hipinfo-less amd-smi reports the gfx target.
            if [ -z "$_gfx_all" ]; then
-                _gfx_all=$(amd-smi static --asic 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}')
+                _gfx_all=$(amd-smi static --asic 2>/dev/null | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+            fi
+        fi
+        # get_torch_index_url reads the arch with ROCR/HIP masks cleared, so a
+        # mask hiding every agent (e.g. ROCR_VISIBLE_DEVICES=-1) still lands
+        # here on a generic rocm index; re-probe unmasked or a masked-out Strix
+        # box keeps the broken generic wheels. Partial masks never get here
+        # (they enumerate at least one agent above) and keep their selection.
+        # ${VAR+x} (not :-): a SET-but-empty mask also hides every agent and
+        # must trigger the re-probe too.
+        if [ -z "$_gfx_all" ] && [ -n "${ROCR_VISIBLE_DEVICES+x}${HIP_VISIBLE_DEVICES+x}" ]; then
+            if command -v rocminfo >/dev/null 2>&1; then
+                _gfx_all=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; rocminfo 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+            fi
+            if [ -z "$_gfx_all" ] && command -v amd-smi >/dev/null 2>&1; then
+                _gfx_all=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; amd-smi list 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
+                [ -z "$_gfx_all" ] && \
+                    _gfx_all=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; amd-smi static --asic 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
            fi
        fi
        _runtime_gfx=""
@ -2863,15 +3298,16 @@ case "$TORCH_INDEX_URL" in
        fi
        _strix_gfx=""
        case "$_runtime_gfx" in
-            gfx1151|gfx1150) _strix_gfx="$_runtime_gfx" ;;
+            gfx1151|gfx1150|gfx1152) _strix_gfx="$_runtime_gfx" ;;
        esac
-        if [ -n "$_strix_gfx" ]; then
+        # Skip rocm7.13+ generic indexes: they already ship the fixes, so the
+        # arch build (rocm7.13) would be a downgrade rather than a rescue.
+        if [ -n "$_strix_gfx" ] && _rocm_leaf_below "$_torch_index_leaf" 7 13; then
            echo "" >&2
-            echo "  [WARN] $_strix_gfx (Strix) + ROCm 7.1 detected -- known _grouped_mm segfault" >&2
-            echo "  [WARN] ROCm 7.1 wheels are broken for gfx1150/gfx1151 (moe_utils.py:167)" >&2
-            echo "  [WARN] Routing to AMD arch-specific index (torch 2.11+rocm7.13 has the real fix)" >&2
-            echo "  [WARN] Upgrade ROCm to 7.2+ to use the standard index:" >&2
-            echo "  [WARN]   https://rocm.docs.amd.com/en/latest/deploy/linux/index.html" >&2
+            echo "  [WARN] $_strix_gfx (Strix) detected -- routing to the AMD arch-specific index" >&2
+            echo "  [WARN] torch 2.11+rocm7.13 has AMD's real gfx1150/gfx1151 fixes (the ROCm 7.1" >&2
+            echo "  [WARN] _grouped_mm segfault, moe_utils.py:167, and later Strix kernel bugs)," >&2
+            echo "  [WARN] and is more reliable than the rocm7.2 index or an offline Radeon repo." >&2
            echo "" >&2
            # AMD's arch-specific index serves torch 2.11.0+rocm7.13.0 which has AMD's
            # actual fix for the gfx1151/gfx1150 _grouped_mm kernel bug -- preferred
@ -2958,12 +3394,14 @@ elif case "$TORCH_INDEX_URL" in */rocm*|*/gfx*) true ;; *) false ;; esac; then
        # gfx1102 matched BEFORE gfx1100 so the spaceless "RX 7700S" lands on
        # gfx1102 (bash case has no negative lookahead like the PS tables).
        case "$_gpu_disp_mkt" in
-            *"9070 XT"*|*9080*)                                                                            _gpu_disp_gfx="gfx1201" ;;  # RDNA 4
-            *9070*|*9060*)                                                                                 _gpu_disp_gfx="gfx1200" ;;  # RDNA 4
-            *"8060S"*|*"8050S"*|*"8040S"*|*"Strix Halo"*|*"Ryzen AI Max"*|*"AI Max"*) _gpu_disp_gfx="gfx1151" ;;  # RDNA 3.5 (Strix Halo: Radeon 8060S/8050S/8040S iGPU, Ryzen AI Max+)
-            *"890M"*|*"880M"*|*"860M"*|*"840M"*|*"Strix Point"*|*"Krackan"*|*"HX 37"*|*"AI 9 HX"*|*"AI 9 36"*|*"AI 7 35"*|*"AI 5 34"*|*"AI 7 PRO 35"*|*"AI 5 33"*) _gpu_disp_gfx="gfx1150" ;;  # RDNA 3.5 (Strix/Krackan Point: Radeon 890M/880M iGPU, Ryzen AI 9 HX 370/375)
-            *"RX 7600"*|*"RX 7700S"*|*"RX 7650"*|*"PRO W7600"*|*"PRO W7500"*|*"PRO V710"*)                  _gpu_disp_gfx="gfx1102" ;;  # RDNA 3 (Navi 33)
-            *"RX 7900"*|*"RX 7800"*|*"RX 7700"*|*"PRO W7900"*|*"PRO W7800"*|*"PRO W7700"*)                  _gpu_disp_gfx="gfx1100" ;;  # RDNA 3 desktop / workstation (Navi 31)
+            *9070*|*9080*)                                                                                 _gpu_disp_gfx="gfx1201" ;;  # RDNA 4 (Navi 48)
+            *9060*)                                                                                        _gpu_disp_gfx="gfx1200" ;;  # RDNA 4 (Navi 44)
+            *"8065S"*|*"8060S"*|*"8050S"*|*"8040S"*|*"Strix Halo"*|*"Ryzen AI Max"*|*"AI Max"*) _gpu_disp_gfx="gfx1151" ;;  # RDNA 3.5 (Strix Halo + Gorgon Halo: Radeon 8065S/8060S/8050S/8040S iGPU, Ryzen AI Max / Max+)
+            *"890M"*|*"880M"*|*"Strix Point"*|*"HX 37"*|*"AI 9 HX"*|*"AI 9 36"*) _gpu_disp_gfx="gfx1150" ;;  # RDNA 3.5 (Strix Point: Radeon 890M/880M, Ryzen AI 9 HX 370/375)
+            *"860M"*|*"840M"*|*"Krackan"*|*"AI 7 35"*|*"AI 5 34"*|*"AI 7 PRO 35"*|*"AI 5 33"*) _gpu_disp_gfx="gfx1152" ;;  # RDNA 3.5 (Krackan Point: Radeon 860M/840M, Ryzen AI 7 350 / AI 5 340)
+            *"RX 7600"*|*"RX 7700S"*|*"RX 7650"*|*"PRO W7600"*|*"PRO W7500"*)                              _gpu_disp_gfx="gfx1102" ;;  # RDNA 3 (Navi 33)
+            *"RX 7800"*|*"RX 7700"*|*"PRO W7700"*|*"PRO V710"*)                                            _gpu_disp_gfx="gfx1101" ;;  # RDNA 3 (Navi 32)
+            *"RX 7900"*|*"PRO W7900"*|*"PRO W7800"*)                                                       _gpu_disp_gfx="gfx1100" ;;  # RDNA 3 desktop / workstation (Navi 31)
            *"780M"*|*"760M"*|*"740M"*|*"Phoenix"*|*"Hawk Point"*|*"Z1 Extreme"*|*"Z2 Extreme"*)            _gpu_disp_gfx="gfx1103" ;;  # RDNA 3 iGPU (Phoenix / Hawk Point)
            *"RX 6900"*|*"RX 6800"*|*"RX 6750"*|*"RX 6700"*|*"PRO W6800"*|*"PRO W6900"*)                    _gpu_disp_gfx="gfx1030" ;;  # RDNA 2 (Navi 21)
            *"RX 6650"*|*"RX 6600"*|*"PRO W6600"*|*"PRO W6650"*)                                            _gpu_disp_gfx="gfx1032" ;;  # RDNA 2 (Navi 23)
@ -2995,6 +3433,17 @@ elif case "$TORCH_INDEX_URL" in */rocm*|*/gfx*) true ;; *) false ;; esac; then
 elif [ "$OS" = "macos" ] && [ "$_ARCH" = "arm64" ]; then
    # Apple Silicon: PyTorch gets Metal (MPS) acceleration over unified memory, so not CPU-only.
    step "gpu" "Apple Silicon (Metal, unified memory)"
+elif _has_amd_rocm_gpu; then
+    if [ "$_torch_index_pinned" = true ]; then
+        # An explicit UNSLOTH_TORCH_INDEX_URL/_FAMILY pin skipped all probing;
+        # do not claim ROCm is unusable when a CPU/other index was requested.
+        step "gpu" "AMD GPU (torch index pinned: $_torch_index_leaf)" "$C_WARN"
+    else
+        # AMD GPU visible to the kernel but the torch index stayed CPU: no usable
+        # ROCm userspace to pick a wheel. "none" would repeat the false diagnosis
+        # this installer used to give.
+        step "gpu" "AMD GPU (no usable ROCm -- CPU fallback)" "$C_WARN"
+    fi
 else
    step "gpu" "none (CPU-only)" "$C_WARN"
 fi
@ -3003,8 +3452,17 @@ fi
 case "$TORCH_INDEX_URL" in
    */cpu)
        if [ "$SKIP_TORCH" = false ] && [ "$OS" != "macos" ]; then
-            substep "No GPU detected -- installing CPU-only PyTorch." "$C_WARN"
-            if [ "$OS" = "wsl" ]; then
+            if [ "$_torch_index_pinned" = true ]; then
+                # An explicit CPU pin is a request, not a detection failure:
+                # skip the SDK guidance (ROCm may be perfectly healthy here).
+                substep "CPU-only PyTorch (index pinned via UNSLOTH_TORCH_INDEX_URL / _FAMILY)."
+            elif _has_amd_rocm_gpu; then
+                substep "AMD GPU detected, but no usable ROCm/HIP install -- installing CPU-only PyTorch." "$C_WARN"
+                substep "Install the ROCm/HIP SDK and re-run this installer for GPU PyTorch." "$C_WARN"
+            else
+                substep "No GPU detected -- installing CPU-only PyTorch." "$C_WARN"
+            fi
+            if [ "$OS" = "wsl" ] && [ "$_torch_index_pinned" = false ]; then
                # WSL + no GPU detected (detection above found nothing). Common
                # cause: an AMD GPU whose ROCm-on-WSL runtime isn't exposed yet --
                # /dev/dxg present (graphics) but no ROCm runtime.
@ -3031,6 +3489,13 @@ case "$TORCH_INDEX_URL" in
                substep "  driver is current; or run unsloth/scripts/install_rocm_wsl_strixhalo.sh yourself."
            else
                substep "AMD ROCm users: see https://docs.unsloth.ai/get-started/install-and-update/amd"
+                # Only when ROCm truly can't see the GPU: a detected-but-too-old
+                # ROCm (rocminfo works, wheels need 6.0+) has its own guidance.
+                if ! _has_amd_rocm_gpu && _amd_gpu_present_via_pci; then
+                    substep "An AMD GPU is on the PCI bus but ROCm cannot see it (no /dev/kfd," "$C_WARN"
+                    substep "  rocminfo, or amd-smi). Install the ROCm kernel stack so /dev/kfd exists;"
+                    substep "  Strix Halo (gfx1151/gfx1150) needs a recent kernel (6.11+) and ROCm 7.x."
+                fi
            fi
            substep "Re-run with --no-torch for GGUF-only (faster, no PyTorch):"
            substep "  curl -fsSL https://unsloth.ai/install.sh | sh -s -- --no-torch"
@ -3096,7 +3561,7 @@ if [ "$_MIGRATED" = true ]; then
        # to prevent transitive torch resolution.
        run_install_cmd_retry "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
            --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4"
+            "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6"
        # Resolve pydantic WITH deps so pip pins pydantic-core to the
        # matching version (no-torch-runtime.txt below is --no-deps).
        # All transitive deps are torch-free.
@ -3113,7 +3578,7 @@ if [ "$_MIGRATED" = true ]; then
        run_install_cmd_retry "install unsloth (migrated)" uv pip install --python "$_VENV_PY" \
            ${_UNSLOTH_TORCH_OVERRIDES:+--overrides "$_UNSLOTH_TORCH_OVERRIDES"} \
            --reinstall-package unsloth --reinstall-package unsloth-zoo \
-            "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" ${_MLX_LM_EXCLUDE_ARG:-}
+            "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" ${_MLX_LM_EXCLUDE_ARG:-}
        [ -n "$_UNSLOTH_TORCH_OVERRIDES" ] && rm -f "$_UNSLOTH_TORCH_OVERRIDES"
        _UNSLOTH_TORCH_OVERRIDES=""
    fi
@ -3337,7 +3802,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
        # runtime deps (typer, safetensors, transformers, etc.) with --no-deps.
        run_install_cmd_retry "install unsloth (no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
            --upgrade-package unsloth --upgrade-package unsloth-zoo \
-            "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4"
+            "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6"
        # Same pydantic-with-deps trick as the migrated branch.
        run_install_cmd_retry "install pydantic (with deps for compatible core)" \
            uv pip install --python "$_VENV_PY" pydantic
@ -3356,7 +3821,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
    elif [ "$STUDIO_LOCAL_INSTALL" = true ]; then
        run_install_cmd_retry "install unsloth (local)" uv pip install --python "$_VENV_PY" \
            ${_UNSLOTH_TORCH_OVERRIDES:+--overrides "$_UNSLOTH_TORCH_OVERRIDES"} \
-            --upgrade-package unsloth "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4"
+            --upgrade-package unsloth "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6"
        substep "overlaying local repo (editable)..."
        run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
        substep "overlaying unsloth-zoo from git main..."
@ -3384,7 +3849,7 @@ else
    tauri_log "STEP" "Installing Unsloth"
    substep "installing unsloth (this may take a few minutes)..."
    if [ "$STUDIO_LOCAL_INSTALL" = true ]; then
-        run_install_cmd_retry "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" "unsloth-zoo>=2026.7.4" "unsloth>=2026.7.4" --torch-backend=auto
+        run_install_cmd_retry "install unsloth (auto torch backend)" uv pip install --python "$_VENV_PY" "unsloth-zoo>=2026.7.6" "unsloth>=2026.7.5" --torch-backend=auto
        substep "overlaying local repo (editable)..."
        run_install_cmd "overlay local repo" uv pip install --python "$_VENV_PY" -e "$_REPO_ROOT" --no-deps
        substep "overlaying unsloth-zoo from git main..."
@ -3396,6 +3861,15 @@ else
    fi
 fi

+_installed_package_version=$("$_VENV_PY" -c \
+    'from importlib.metadata import version; import sys; print(version(sys.argv[1]))' \
+    "$PACKAGE_NAME" 2>/dev/null || true)
+if [ -n "$_installed_package_version" ]; then
+    step "$PACKAGE_NAME" "$_installed_package_version installed"
+else
+    substep "[WARN] installed $PACKAGE_NAME version could not be determined" "$C_WARN"
+fi
+
 # ── Enforce the installed torch flavor matches the detected GPU build ──
 # PEP 440 ignores the +cpu/+cuXXX/+rocm local label in a version range, so uv
 # keeps a stale torch==X+cpu against a GPU index and the venv silently trains on
@ -3629,9 +4103,11 @@ echo ""
 # In non-interactive environments (Docker, CI, cloud-init) just print instructions.
 if [ "$_SKIP_AUTOSTART" != true ] && [ -t 1 ]; then
    echo ""
-    printf "  Start Unsloth Studio now? [Y/n] "
    # No readable answer (closed/EOF tty) defaults to no; Enter is still yes.
-    if [ -r /dev/tty ]; then
+    # Prompt only when something can answer: `test -r` passes on the unopenable
+    # /dev/tty found in containers, leaving a dangling question in the log.
+    if _can_read_tty; then
+        printf "  Start Unsloth Studio now? [Y/n] "
        read -r _reply </dev/tty || _reply="n"
    else
        _reply="n"
--- a/pyproject.toml
+++ b/pyproject.toml
@ -25,7 +25,7 @@ classifiers = [
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
 ]
 dependencies = [
-    "typer",
+    "typer>=0.12.0",
    "rich",
    "pydantic",
    "pyyaml",
@ -42,7 +42,7 @@ version = {attr = "unsloth.models._utils.__version__"}
 include-package-data = true

 [tool.setuptools.package-data]
-unsloth_cli = ["codex_fallback_prompt.md"]
+unsloth_cli = ["codex_fallback_prompt.md", "pi_subagent.ts"]
 studio = [
    "*.sh",
    "*.ps1",
@ -74,7 +74,7 @@ triton = [
 ]

 huggingfacenotorch = [
-    "unsloth_zoo>=2026.7.4",
+    "unsloth_zoo>=2026.7.6",
    "wheel>=0.42.0",
    "packaging",
    "numpy",
@ -93,9 +93,20 @@ huggingfacenotorch = [
    "trl>=0.18.2,!=0.19.0,<=0.24.0",
    "sentence-transformers",
 ]
+# torchcodec backend for Gemma audio / datasets>=4 (#7225).
+# Pick the audio-torch* pin matching your torch minor (see TORCH_TORCHCODEC).
+audio-torch210 = [
+    "torchcodec>=0.10.0,<0.11.0 ; python_version >= '3.10'",
+]
+audio-torch290 = [
+    "torchcodec>=0.8.0,<0.10.0 ; python_version >= '3.10'",
+]
+audio-torch280 = [
+    "torchcodec>=0.6.0,<0.8.0 ; python_version >= '3.9'",
+]
 huggingface = [
    "unsloth[huggingfacenotorch]",
-    "unsloth_zoo>=2026.7.4",
+    "unsloth_zoo>=2026.7.6",
    "torchvision",
    "unsloth[triton]",
 ]
@ -564,16 +575,19 @@ cu126-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu126onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu128-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu128onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu130-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu130onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu126-torch2110 = [
    "unsloth[huggingface]",
@ -627,7 +641,7 @@ colab-ampere-torch220 = [
    "flash-attn>=2.6.3 ; ('linux' in sys_platform)",
 ]
 colab-new = [
-    "unsloth_zoo>=2026.7.4",
+    "unsloth_zoo>=2026.7.6",
    "packaging",
    "tyro",
    "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
@ -878,16 +892,19 @@ cu126-ampere-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu126onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu128-ampere-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu128onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu130-ampere-torch2100 = [
    "unsloth[huggingface]",
    "bitsandbytes>=0.45.5,!=0.46.0,!=0.48.0",
    "unsloth[cu130onlytorch2100]",
+    "unsloth[audio-torch210]",
 ]
 cu126-ampere-torch2110 = [
    "unsloth[huggingface]",
@ -1187,7 +1204,8 @@ intelgputorch210 = [
    "torchvision @ https://download.pytorch.org/whl/xpu/torchvision-0.25.0%2Bxpu-cp313-cp313-win_amd64.whl#sha256=1c4b44b36a557f7381e3076fb8843366742238648441d607c8d049c6da0f8886 ; sys_platform == 'win32' and python_version == '3.13' and (platform_machine == 'AMD64' or platform_machine == 'x86_64')",
 ]
 intel-gpu-torch210 = [
-    "unsloth[intelgputorch210]"
+    "unsloth[intelgputorch210]",
+    "unsloth[audio-torch210]",
 ]
 intelgputorch2110 = [
    "unsloth_zoo[intelgpu]",
@ -1341,6 +1359,7 @@ rocm72-torch2100 = [
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.2/torchvision-0.25.0%2Brocm7.2.0.git82df5f59-cp311-cp311-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.11' and platform_machine == 'x86_64'",
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.2/torchvision-0.25.0%2Brocm7.2.0.git82df5f59-cp312-cp312-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.12' and platform_machine == 'x86_64'",
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.2/torchvision-0.25.0%2Brocm7.2.0.git82df5f59-cp313-cp313-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.13' and platform_machine == 'x86_64'",
+    "unsloth[audio-torch210]",
 ]
 rocm711-torch2100 = [
    "unsloth[amd]",
@ -1359,6 +1378,7 @@ rocm711-torch2100 = [
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.1.1/torchvision-0.25.0%2Brocm7.1.1.git82df5f59-cp311-cp311-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.11' and platform_machine == 'x86_64'",
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.1.1/torchvision-0.25.0%2Brocm7.1.1.git82df5f59-cp312-cp312-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.12' and platform_machine == 'x86_64'",
    "torchvision @ https://repo.radeon.com/rocm/manylinux/rocm-rel-7.1.1/torchvision-0.25.0%2Brocm7.1.1.git82df5f59-cp313-cp313-linux_x86_64.whl ; platform_system == 'Linux' and python_version == '3.13' and platform_machine == 'x86_64'",
+    "unsloth[audio-torch210]",
 ]

 [project.urls]
--- a/scripts/build_whisper_cpp.sh
+++ b/scripts/build_whisper_cpp.sh
@ -0,0 +1,71 @@
+#!/bin/sh
+# Build whisper.cpp's whisper-server for Studio's GGUF dictation engine.
+#
+# Installs into the managed Studio home so the backend's binary discovery
+# (core/inference/stt_ggml_sidecar.py::find_whisper_server_binary) picks it up:
+#   <UNSLOTH_STUDIO_HOME>/whisper.cpp/build/bin/whisper-server   (custom home)
+#   ~/.unsloth/whisper.cpp/build/bin/whisper-server              (default)
+#
+# Usage:
+#   ./scripts/build_whisper_cpp.sh              # build the pinned tag
+#   WHISPER_CPP_TAG=v1.9.0 ./scripts/build_whisper_cpp.sh
+#
+# Requires: git, cmake, a C/C++ toolchain (the same prerequisites as a
+# llama.cpp source build). GPU backends are auto-detected by whisper.cpp's
+# CMake (Metal on macOS; set GGML_CUDA=1 to force a CUDA build on Linux).
+
+set -eu
+
+WHISPER_CPP_SOURCE="${WHISPER_CPP_SOURCE:-https://github.com/ggml-org/whisper.cpp}"
+WHISPER_CPP_TAG="${WHISPER_CPP_TAG:-v1.9.1}"
+
+STUDIO_HOME="${UNSLOTH_STUDIO_HOME:-${STUDIO_HOME:-}}"
+CUSTOM_STUDIO_HOME=false
+if [ -n "$STUDIO_HOME" ]; then
+    CUSTOM_STUDIO_HOME=true
+    INSTALL_DIR="$STUDIO_HOME/whisper.cpp"
+else
+    INSTALL_DIR="$HOME/.unsloth/whisper.cpp"
+fi
+
+command -v git >/dev/null 2>&1 || { echo "ERROR: git is required" >&2; exit 1; }
+command -v cmake >/dev/null 2>&1 || { echo "ERROR: cmake is required" >&2; exit 1; }
+
+# Same policy as studio/setup.sh's _assert_studio_owned_or_absent: never delete
+# a directory under a custom Studio home unless Studio itself created it (the
+# marker file below). Protects a user-managed whisper.cpp/src from rm -rf.
+STUDIO_OWNED_MARKER=".unsloth-studio-owned"
+if [ "$CUSTOM_STUDIO_HOME" = true ] && [ -e "$INSTALL_DIR" ] && \
+   [ ! -f "$INSTALL_DIR/$STUDIO_OWNED_MARKER" ]; then
+    echo "ERROR: $INSTALL_DIR already exists and is not marked as an Unsloth-owned whisper.cpp build tree." >&2
+    echo "       Move it aside or choose an empty UNSLOTH_STUDIO_HOME before re-running." >&2
+    exit 1
+fi
+
+echo "==> Building whisper.cpp ($WHISPER_CPP_TAG) into $INSTALL_DIR"
+mkdir -p "$INSTALL_DIR"
+: > "$INSTALL_DIR/$STUDIO_OWNED_MARKER"
+
+if [ ! -d "$INSTALL_DIR/src/.git" ]; then
+    rm -rf "$INSTALL_DIR/src"
+    git clone --depth 1 --branch "$WHISPER_CPP_TAG" "$WHISPER_CPP_SOURCE" "$INSTALL_DIR/src"
+else
+    git -C "$INSTALL_DIR/src" fetch --depth 1 origin "$WHISPER_CPP_TAG"
+    git -C "$INSTALL_DIR/src" checkout FETCH_HEAD
+fi
+
+CMAKE_FLAGS="-DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=OFF"
+if [ "${GGML_CUDA:-0}" = "1" ]; then
+    CMAKE_FLAGS="$CMAKE_FLAGS -DGGML_CUDA=ON"
+fi
+
+# shellcheck disable=SC2086
+cmake -S "$INSTALL_DIR/src" -B "$INSTALL_DIR/src/build" $CMAKE_FLAGS
+NCPU="$(getconf _NPROCESSORS_ONLN 2>/dev/null || echo 4)"
+cmake --build "$INSTALL_DIR/src/build" --config Release --target whisper-server -j"$NCPU"
+
+mkdir -p "$INSTALL_DIR/build/bin"
+cp "$INSTALL_DIR/src/build/bin/whisper-server" "$INSTALL_DIR/build/bin/whisper-server"
+
+echo "==> Installed $INSTALL_DIR/build/bin/whisper-server"
+"$INSTALL_DIR/build/bin/whisper-server" --help >/dev/null 2>&1 && echo "==> Binary runs OK"
--- a/scripts/notebook_validator.py
+++ b/scripts/notebook_validator.py
@ -95,8 +95,8 @@ COLAB_ORACLE_BASE_URL = "https://raw.githubusercontent.com/googlecolab/backend-i
 # Source: pytorch/torchcodec compatibility matrix on its README.
 TORCH_TORCHCODEC: dict[str, set[str]] = {
    "2.10": {"0.10"},
-    "2.9": {"0.7", "0.8", "0.9"},
-    "2.8": {"0.6"},
+    "2.9": {"0.8", "0.9"},
+    "2.8": {"0.6", "0.7"},
    "2.7": {"0.3", "0.4", "0.5"},
    "2.6": {"0.2", "0.3"},
    "2.5": {"0.1", "0.2"},
--- a/scripts/scan_packages_baseline.json
+++ b/scripts/scan_packages_baseline.json
@ -1,5 +1,5 @@
 {
-  "_comment": "scan_packages.py allowlist. Each entry is a CRITICAL/HIGH finding manually judged benign. Matched on (package, package-relative file, check, evidence_hash); evidence_hash is over the matched code with L<NN>: markers stripped, so version bumps and line shifts do not reopen an entry but changed code does. severity and evidence are for review only. Regenerate with --write-baseline AFTER reviewing every line.",
+  "_comment": "scan_packages.py allowlist (reviewed). Each entry is a CRITICAL/HIGH finding manually judged benign. Matched on (package, package-relative file, check, evidence_hash); evidence_hash is over the matched code with L<NN>: markers stripped, so version bumps and line shifts do not reopen an entry but changed code does. severity and evidence are for review only. Regenerate with --write-baseline AFTER reviewing every line.",
  "version": 1,
  "entries": [
    {
@ -303,8 +303,8 @@
      "file": "openai/_base_client.py",
      "check": "C2 polling/beaconing loop detected",
      "severity": "CRITICAL",
-      "evidence": "L264:         while True: sha256:95ca67e46d42354ae650abbdc5b0d97df8b0ed43187800bf40f5690c3901b94b",
-      "evidence_hash": "a57d8d15fed0bf04f9967dcc18a18b80bb19f4095675bccbb78ac0450d7fce14"
+      "evidence": "L274:         while True: sha256:90a38e5c1e26893c7c273354143612640e9a9c0f079d3e2b60612d79f24e80a6",
+      "evidence_hash": "1022e8e8649436ec64a98a9d9141d085452c49549fd2157b0278fc369a83ac66"
    },
    {
      "package": "openai",
@ -319,8 +319,8 @@
      "file": "openai/auth/_workload.py",
      "check": "Accesses cloud metadata/IMDS AND makes network calls",
      "severity": "CRITICAL",
-      "evidence": "IMDS: L96:             url = \"http://169.254.169.254/metadata/identity/oauth2/token\" | L149:             url = \"http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/identity\"\nNetwork: L77:     http_client: httpx.Client | None = None, | L108:                 with httpx.Client() as client: | L133:     http_client: httpx.Client | None = None, | L155:                 with httpx.Client() as client: | L248:         with httpx.Client() as client:",
-      "evidence_hash": "1581d9f4a23393e9af23fbe5ef9f66807b22c5b5a3f1fe167254c9ebee108567"
+      "evidence": "IMDS: L97:             url = \"http://169.254.169.254/metadata/identity/oauth2/token\" | L150:             url = \"http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/identity\"\nNetwork: L78:     http_client: httpx.Client | None = None, | L109:                 with httpx.Client() as client: | L134:     http_client: httpx.Client | None = None, | L156:                 with httpx.Client() as client: | L251:         exchange_client = DefaultHttpx2Client(follow_redirects=False) if self._use_httpx2 else httpx.Client()",
+      "evidence_hash": "9717e51cb961dc14c458955d91a1e48e3753997346ecea0106bded3a8d64bfe0"
    },
    {
      "package": "openai",
@ -343,8 +343,8 @@
      "file": "openai/resources/beta/responses/responses.py",
      "check": "C2 polling/beaconing loop detected",
      "severity": "CRITICAL",
-      "evidence": "L3999:         while True: sha256:df298b6eaf3416589b79f4ef283f8fb76e54d505bfda8840673f8e6419117e2e",
-      "evidence_hash": "10ce5cb5a7097fcff4042ddcfb4802edda60aa4b7b113c8b926a52ddb76f78c2"
+      "evidence": "L4000:         while True: sha256:f8ab538118daba9ec06e27399dbdc90a4521c3390e6a47a6348a1f180a83effd",
+      "evidence_hash": "31481ea83c687acc27144d72d3832d4fb98dd1c79fb5e0ddd85080de95997b9f"
    },
    {
      "package": "openai",
@ -359,16 +359,16 @@
      "file": "openai/resources/realtime/realtime.py",
      "check": "C2 polling/beaconing loop detected",
      "severity": "CRITICAL",
-      "evidence": "L310:         while True: sha256:458198ff3d3f05870bf98c9564cbfd68c739e57b9bbe4120ed81e3eb6af74a05",
-      "evidence_hash": "a3165d21e46b3ce553795daeae53e8f80e8e89c5cb228e68e6dcaff54bca5a89"
+      "evidence": "L311:         while True: sha256:5b63313072aae9ca28677e03426513ccf12221e4f4e0ea6c31efbe09790633b5",
+      "evidence_hash": "05e1af469d651b51673763a7c4cdf759af9472fb627b7b470adc28cc237bd650"
    },
    {
      "package": "openai",
      "file": "openai/resources/responses/responses.py",
      "check": "C2 polling/beaconing loop detected",
      "severity": "CRITICAL",
-      "evidence": "L3950:         while True: sha256:1ce0b5a388c747945cdfda1a71b77afdfd03ae840d7aa9fa62f02eb00aa5e29f",
-      "evidence_hash": "6de300ebb5e6e17cb51c89cbcdf08515a44655182f0776f0908a9d1043ebbcd7"
+      "evidence": "L3951:         while True: sha256:d68ef896bf0743ca430cfacb9a3353da1f3b9c51c3a21b6450a07a32b55aa2ac",
+      "evidence_hash": "160eecdd79b521bffbe8476f782b69a0724c35d1b19376a7600807165fd54f9f"
    },
    {
      "package": "openai",
@ -1545,6 +1545,78 @@
      "severity": "HIGH",
      "evidence": "Obfusc: L836:         code = compile(module, \"<werkzeug routing>\", \"exec\")\nExec: L736:         exec(code, globs, locs)",
      "evidence_hash": "5c0992c90f05c772abd94d00784f157de337e1f8567f8b3aee1b15e46c96cd5d"
+    },
+    {
+      "package": "unsloth-zoo",
+      "file": "tests/test_mlx_save_export_regressions.py",
+      "check": "Writes to /tmp and executes (staged dropper)",
+      "severity": "CRITICAL",
+      "evidence": "L165:         temporary_location=\"/tmp/ignored\", sha256:ab5c587f9ec31a0cc10ee55698ab133a417148d9d3f371bbc81b1e13fa119c13",
+      "evidence_hash": "93a11159147aad94f353ec4d2e0b8486b256abef88cd96d741813222cd32b138"
+    },
+    {
+      "package": "unsloth-zoo",
+      "file": "tests/test_vision_collator_audio.py",
+      "check": "Writes to /tmp and executes (staged dropper)",
+      "severity": "CRITICAL",
+      "evidence": "L111:         out = extract_audio_info(msgs({\"type\": \"audio\", key: \"/tmp/a.wav\"})) sha256:2efe23ffbe2b91b8403aec9b700736919b59e5ca770f8e1f5501651b44b7d398",
+      "evidence_hash": "d416b79dd17b24214f3f7653ac01354507d7bf0fc464dee30a4a4b8998f063ba"
+    },
+    {
+      "package": "openai",
+      "file": "openai/_base_client.py",
+      "check": "C2 polling/beaconing loop detected",
+      "severity": "CRITICAL",
+      "evidence": "L274:         while True: sha256:90a38e5c1e26893c7c273354143612640e9a9c0f079d3e2b60612d79f24e80a6",
+      "evidence_hash": "1022e8e8649436ec64a98a9d9141d085452c49549fd2157b0278fc369a83ac66"
+    },
+    {
+      "package": "openai",
+      "file": "openai/auth/_workload.py",
+      "check": "Accesses cloud metadata/IMDS AND makes network calls",
+      "severity": "CRITICAL",
+      "evidence": "IMDS: L97:             url = \"http://169.254.169.254/metadata/identity/oauth2/token\" | L150:             url = \"http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/identity\"\nNetwork: L78:     http_client: httpx.Client | None = None, | L109:                 with httpx.Client() as client: | L134:     http_client: httpx.Client | None = None, | L156:                 with httpx.Client() as client: | L251:         exchange_client = DefaultHttpx2Client(follow_redirects=False) if self._use_httpx2 else httpx.Client()",
+      "evidence_hash": "9717e51cb961dc14c458955d91a1e48e3753997346ecea0106bded3a8d64bfe0"
+    },
+    {
+      "package": "openai",
+      "file": "openai/resources/beta/responses/responses.py",
+      "check": "C2 polling/beaconing loop detected",
+      "severity": "CRITICAL",
+      "evidence": "L4000:         while True: sha256:f8ab538118daba9ec06e27399dbdc90a4521c3390e6a47a6348a1f180a83effd",
+      "evidence_hash": "31481ea83c687acc27144d72d3832d4fb98dd1c79fb5e0ddd85080de95997b9f"
+    },
+    {
+      "package": "openai",
+      "file": "openai/resources/realtime/realtime.py",
+      "check": "C2 polling/beaconing loop detected",
+      "severity": "CRITICAL",
+      "evidence": "L311:         while True: sha256:5b63313072aae9ca28677e03426513ccf12221e4f4e0ea6c31efbe09790633b5",
+      "evidence_hash": "05e1af469d651b51673763a7c4cdf759af9472fb627b7b470adc28cc237bd650"
+    },
+    {
+      "package": "openai",
+      "file": "openai/resources/responses/responses.py",
+      "check": "C2 polling/beaconing loop detected",
+      "severity": "CRITICAL",
+      "evidence": "L3951:         while True: sha256:d68ef896bf0743ca430cfacb9a3353da1f3b9c51c3a21b6450a07a32b55aa2ac",
+      "evidence_hash": "160eecdd79b521bffbe8476f782b69a0724c35d1b19376a7600807165fd54f9f"
+    },
+    {
+      "package": "unsloth-zoo",
+      "file": "tests/test_gemma4_forced_float32_ple_dtype.py",
+      "check": "Advanced obfuscation (marshal/compile/zlib) + exec/eval",
+      "severity": "HIGH",
+      "evidence": "Obfusc: L277:     compile(rewritten + _GEMMA4_PLE_CAST_HELPER, \"<gemma4-ple-generated>\", \"exec\") | L440:     compile(on, \"<gemma4-ple-append>\", \"exec\") | L468:     compile(generated, \"<gemma4-ple-crosspath>\", \"exec\")\nExec: L19:     exec(_GEMMA4_PLE_CAST_HELPER, namespace)",
+      "evidence_hash": "a85e24d8e7c431563cbd83b70f91a3b971abde0f37083d68e70984147960cc70"
+    },
+    {
+      "package": "unsloth-zoo",
+      "file": "tests/test_vision_collator_audio.py",
+      "check": "Writes to /tmp and executes (staged dropper)",
+      "severity": "CRITICAL",
+      "evidence": "L111:         out = extract_audio_info(msgs({\"type\": \"audio\", key: \"/tmp/a.wav\"})) sha256:022f81dd21acfc6a35a058de96132834c218404a9e37b3d09a7768a8c8f6c728",
+      "evidence_hash": "2d1e75446af120d9133a42aa8af426a839d3434d9dc109cc1d6c1b22ca1ddb75"
    }
  ]
 }
--- a/studio/Unsloth_Studio_Colab.ipynb
+++ b/studio/Unsloth_Studio_Colab.ipynb
@ -1,134 +1,145 @@
 {
- "cells": [
-  {
-   "cell_type": "markdown",
-   "metadata": {
-    "id": "view-in-github",
-    "colab_type": "text"
-   },
-   "source": [
-    "<a href=\"https://colab.research.google.com/github/unslothai/unsloth/blob/main/studio/Unsloth_Studio_Colab.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
-   ]
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "view-in-github",
+        "colab_type": "text"
+      },
+      "source": [
+        "<a href=\"https://colab.research.google.com/github/unslothai/unsloth/blob/main/studio/Unsloth_Studio_Colab.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "6b87de59"
+      },
+      "source": [
+        "To run this, press \"*Runtime*\" and press \"*Run all*\" on a **free** Tesla T4 Google Colab instance!\n",
+        "<div class=\"align-center\">\n",
+        "<a href=\"https://unsloth.ai/\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png\" width=\"115\"></a>\n",
+        "<a href=\"https://discord.gg/unsloth\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/Discord button.png\" width=\"145\"></a>\n",
+        "<a href=\"https://unsloth.ai/docs/\"><img src=\"https://github.com/unslothai/unsloth/blob/main/images/documentation%20green%20button.png?raw=true\" width=\"125\"></a> Join Discord if you need help + ⭐ <i>Star us on <a href=\"https://github.com/unslothai/unsloth\">Github</a> </i> ⭐\n",
+        "</div>\n",
+        "\n",
+        "To install Unsloth Studio on your local device, follow [our guide](https://unsloth.ai/docs/new/unsloth-studio/install). Unsloth Studio is licensed [AGPL-3.0](https://github.com/unslothai/unsloth/blob/main/studio/LICENSE.AGPL-3.0).\n",
+        "\n",
+        "### Unsloth Studio\n",
+        "\n",
+        "Train and run open models with [**Unsloth Studio**](https://unsloth.ai/docs/new/unsloth-studio/start). NEW! Installation should now only take 2 mins!\n",
+        "\n",
+        "\n",
+        "We are actively working on making Unsloth Studio install on Colab T4 GPUs faster.\n",
+        "\n",
+        "[Features](https://unsloth.ai/docs/new/unsloth-studio#features) • [Quickstart](https://unsloth.ai/docs/new/unsloth-studio/start) • [Data Recipes](https://unsloth.ai/docs/new/unsloth-studio/data-recipe) • [Unsloth Chat](https://unsloth.ai/docs/new/unsloth-studio/chat) • [Export](https://unsloth.ai/docs/new/unsloth-studio/export)"
+      ],
+      "id": "6b87de59"
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "e4206349"
+      },
+      "source": [
+        "<p align=\"left\"><img src=\"https://github.com/unslothai/unsloth/raw/main/studio/frontend/public/studio%20github%20landscape%20colab%20display.png\" width=\"600\"></p>"
+      ],
+      "id": "e4206349"
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "27da2957"
+      },
+      "source": [
+        "### Setup: Clone repo and run setup"
+      ],
+      "id": "27da2957"
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "27e68f91"
+      },
+      "source": "!git clone --depth 1 --branch main https://github.com/unslothai/unsloth.git\n%cd /content/unsloth\n!chmod +x studio/setup.sh && ./studio/setup.sh --local",
+      "execution_count": null,
+      "outputs": [],
+      "id": "27e68f91"
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "3e1771a9"
+      },
+      "source": [
+        "### Start Unsloth Studio"
+      ],
+      "id": "3e1771a9"
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "277e431e"
+      },
+      "source": [
+        "import sys\n",
+        "sys.path.insert(0, \"/content/unsloth/studio/backend\")\n",
+        "from colab import start\n",
+        "\n",
+        "# On Colab, start() auto-opens a Cloudflare link and prints admin login credentials.\n",
+        "# Use the Cloudflare link above the ready card to open Studio (in-cell iframes often stay blank).\n",
+        "start()\n",
+        "\n",
+        "# To skip the Cloudflare tunnel and try the in-notebook proxy iframe only:\n",
+        "# start(cloudflare=False)"
+      ],
+      "execution_count": null,
+      "outputs": [],
+      "id": "277e431e"
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "f2b0c6a1"
+      },
+      "source": [
+        "And we're done! If you have any questions on Unsloth, we have a [Discord](https://discord.gg/unsloth) channel! If you find any bugs or want to keep updated with the latest LLM stuff, or need help, join projects etc, feel free to join our Discord!\n",
+        "\n",
+        "Some other resources:\n",
+        "1. Looking to use Unsloth locally? Read our [Installation Guide](https://unsloth.ai/docs/get-started/install) for details on installing Unsloth on Windows, Docker, AMD, Intel GPUs.\n",
+        "2. Learn how to do Reinforcement Learning with our [RL Guide and notebooks](https://unsloth.ai/docs/get-started/reinforcement-learning-rl-guide).\n",
+        "3. Read our guides and notebooks for [Text-to-speech (TTS)](https://unsloth.ai/docs/basics/text-to-speech-tts-fine-tuning) and [vision](https://unsloth.ai/docs/basics/vision-fine-tuning) model support.\n",
+        "4. Explore our [LLM Tutorials Directory](https://unsloth.ai/docs/models/tutorials-how-to-fine-tune-and-run-llms) to find dedicated guides for each model.\n",
+        "5. Need help with Inference? Read our [Inference & Deployment page](https://unsloth.ai/docs/basics/inference-and-deployment) for details on using vLLM, llama.cpp, Ollama etc.\n",
+        "\n",
+        "<div class=\"align-center\">\n",
+        "  <a href=\"https://unsloth.ai\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png\" width=\"115\"></a>\n",
+        "  <a href=\"https://discord.gg/unsloth\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/Discord.png\" width=\"145\"></a>\n",
+        "  <a href=\"https://unsloth.ai/docs/\"><img src=\"https://github.com/unslothai/unsloth/blob/main/images/documentation%20green%20button.png?raw=true\" width=\"125\"></a>\n",
+        "\n",
+        "  Join Discord if you need help + ⭐️ <i>Star us on <a href=\"https://github.com/unslothai/unsloth\">Github</a> </i> ⭐️\n",
+        "\n",
+        "  <b>This notebook is licensed <a href=\"https://github.com/unslothai/unsloth/blob/main/studio/LICENSE.AGPL-3.0\">AGPL-3.0</a></b>\n",
+        "</div>"
+      ],
+      "id": "f2b0c6a1"
+    }
+  ],
+  "metadata": {
+    "accelerator": "GPU",
+    "colab": {
+      "gpuType": "T4",
+      "provenance": [],
+      "include_colab_link": true
+    },
+    "kernelspec": {
+      "display_name": "Python 3",
+      "name": "python3"
+    },
+    "language_info": {
+      "name": "python"
+    }
  },
-  {
-   "cell_type": "markdown",
-   "id": "6b87de59",
-   "metadata": {
-    "id": "6b87de59"
-   },
-   "source": [
-    "To run this, press \"*Runtime*\" and press \"*Run all*\" on a **free** Tesla T4 Google Colab instance!\n",
-    "<div class=\"align-center\">\n",
-    "<a href=\"https://unsloth.ai/\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png\" width=\"115\"></a>\n",
-    "<a href=\"https://discord.gg/unsloth\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/Discord button.png\" width=\"145\"></a>\n",
-    "<a href=\"https://unsloth.ai/docs/\"><img src=\"https://github.com/unslothai/unsloth/blob/main/images/documentation%20green%20button.png?raw=true\" width=\"125\"></a> Join Discord if you need help + ⭐ <i>Star us on <a href=\"https://github.com/unslothai/unsloth\">Github</a> </i> ⭐\n",
-    "</div>\n",
-    "\n",
-    "To install Unsloth Studio on your local device, follow [our guide](https://unsloth.ai/docs/new/unsloth-studio/install). Unsloth Studio is licensed [AGPL-3.0](https://github.com/unslothai/unsloth/blob/main/studio/LICENSE.AGPL-3.0).\n",
-    "\n",
-    "### Unsloth Studio\n",
-    "\n",
-    "Train and run open models with [**Unsloth Studio**](https://unsloth.ai/docs/new/unsloth-studio/start). NEW! Installation should now only take 2 mins!\n",
-    "\n",
-    "\n",
-    "We are actively working on making Unsloth Studio install on Colab T4 GPUs faster.\n",
-    "\n",
-    "[Features](https://unsloth.ai/docs/new/unsloth-studio#features) • [Quickstart](https://unsloth.ai/docs/new/unsloth-studio/start) • [Data Recipes](https://unsloth.ai/docs/new/unsloth-studio/data-recipe) • [Unsloth Chat](https://unsloth.ai/docs/new/unsloth-studio/chat) • [Export](https://unsloth.ai/docs/new/unsloth-studio/export)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "e4206349",
-   "metadata": {
-    "id": "e4206349"
-   },
-   "source": [
-    "<p align=\"left\"><img src=\"https://github.com/unslothai/unsloth/raw/main/studio/frontend/public/studio%20github%20landscape%20colab%20display.png\" width=\"600\"></p>"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "27da2957",
-   "metadata": {
-    "id": "27da2957"
-   },
-   "source": [
-    "### Setup: Clone repo and run setup"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "27e68f91",
-   "metadata": {
-    "id": "27e68f91"
-   },
-   "outputs": [],
-   "source": "!git clone --depth 1 --branch main https://github.com/unslothai/unsloth.git\n%cd /content/unsloth\n!chmod +x studio/setup.sh && ./studio/setup.sh --local"
-  },
-  {
-   "cell_type": "markdown",
-   "id": "3e1771a9",
-   "metadata": {
-    "id": "3e1771a9"
-   },
-   "source": [
-    "### Start Unsloth Studio"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "277e431e",
-   "metadata": {
-    "id": "277e431e"
-   },
-   "outputs": [],
-   "source": "import sys\nsys.path.insert(0, \"/content/unsloth/studio/backend\")\nfrom colab import start\n\n# Default: in-tab iframe only. start() blocks to keep the kernel alive.\nstart()\n\n# For a shareable Cloudflare link, replace start() above with:\n# start(cloudflare=True)"
-  },
-  {
-   "cell_type": "markdown",
-   "id": "f2b0c6a1",
-   "metadata": {
-    "id": "f2b0c6a1"
-   },
-   "source": [
-    "And we're done! If you have any questions on Unsloth, we have a [Discord](https://discord.gg/unsloth) channel! If you find any bugs or want to keep updated with the latest LLM stuff, or need help, join projects etc, feel free to join our Discord!\n",
-    "\n",
-    "Some other resources:\n",
-    "1. Looking to use Unsloth locally? Read our [Installation Guide](https://unsloth.ai/docs/get-started/install) for details on installing Unsloth on Windows, Docker, AMD, Intel GPUs.\n",
-    "2. Learn how to do Reinforcement Learning with our [RL Guide and notebooks](https://unsloth.ai/docs/get-started/reinforcement-learning-rl-guide).\n",
-    "3. Read our guides and notebooks for [Text-to-speech (TTS)](https://unsloth.ai/docs/basics/text-to-speech-tts-fine-tuning) and [vision](https://unsloth.ai/docs/basics/vision-fine-tuning) model support.\n",
-    "4. Explore our [LLM Tutorials Directory](https://unsloth.ai/docs/models/tutorials-how-to-fine-tune-and-run-llms) to find dedicated guides for each model.\n",
-    "5. Need help with Inference? Read our [Inference & Deployment page](https://unsloth.ai/docs/basics/inference-and-deployment) for details on using vLLM, llama.cpp, Ollama etc.\n",
-    "\n",
-    "<div class=\"align-center\">\n",
-    "  <a href=\"https://unsloth.ai\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png\" width=\"115\"></a>\n",
-    "  <a href=\"https://discord.gg/unsloth\"><img src=\"https://github.com/unslothai/unsloth/raw/main/images/Discord.png\" width=\"145\"></a>\n",
-    "  <a href=\"https://unsloth.ai/docs/\"><img src=\"https://github.com/unslothai/unsloth/blob/main/images/documentation%20green%20button.png?raw=true\" width=\"125\"></a>\n",
-    "\n",
-    "  Join Discord if you need help + ⭐️ <i>Star us on <a href=\"https://github.com/unslothai/unsloth\">Github</a> </i> ⭐️\n",
-    "\n",
-    "  <b>This notebook is licensed <a href=\"https://github.com/unslothai/unsloth/blob/main/studio/LICENSE.AGPL-3.0\">AGPL-3.0</a></b>\n",
-    "</div>"
-   ]
-  }
- ],
- "metadata": {
-  "accelerator": "GPU",
-  "colab": {
-   "gpuType": "T4",
-   "provenance": [],
-   "include_colab_link": true
-  },
-  "kernelspec": {
-   "display_name": "Python 3",
-   "name": "python3"
-  },
-  "language_info": {
-   "name": "python"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 5
+  "nbformat": 4,
+  "nbformat_minor": 5
 }
--- a/studio/backend/assets/configs/full_finetune.yaml
+++ b/studio/backend/assets/configs/full_finetune.yaml
@ -30,6 +30,7 @@ lora:
  vision_all_linear: false
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/lora_text.yaml
+++ b/studio/backend/assets/configs/lora_text.yaml
@ -30,6 +30,7 @@ lora:
  vision_all_linear: false
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/default.yaml
+++ b/studio/backend/assets/configs/model_defaults/default.yaml
@ -33,6 +33,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/embedding/unsloth_Qwen3-Embedding-0.6B.yaml
+++ b/studio/backend/assets/configs/model_defaults/embedding/unsloth_Qwen3-Embedding-0.6B.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/embedding/unsloth_all-MiniLM-L6-v2.yaml
+++ b/studio/backend/assets/configs/model_defaults/embedding/unsloth_all-MiniLM-L6-v2.yaml
@ -30,6 +30,7 @@ lora:
    - "query"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/embedding/unsloth_bge-m3.yaml
+++ b/studio/backend/assets/configs/model_defaults/embedding/unsloth_bge-m3.yaml
@ -30,6 +30,7 @@ lora:
    - "value"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/embedding/unsloth_embeddinggemma-300m.yaml
+++ b/studio/backend/assets/configs/model_defaults/embedding/unsloth_embeddinggemma-300m.yaml
@ -33,6 +33,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/embedding/unsloth_gte-modernbert-base.yaml
+++ b/studio/backend/assets/configs/model_defaults/embedding/unsloth_gte-modernbert-base.yaml
@ -29,6 +29,7 @@ lora:
    - "Wqkv"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml
+++ b/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml
+++ b/studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml
@ -33,6 +33,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml
+++ b/studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml
@ -26,6 +26,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml
+++ b/studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml
@ -37,6 +37,7 @@ lora:
    - "shared_mlp.output_linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml
+++ b/studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml
@ -37,6 +37,7 @@ lora:
    - "shared_mlp.output_linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
+++ b/studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
@ -30,6 +30,7 @@ lora:
    - "v_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: false
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml
@ -33,6 +33,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml
@ -33,6 +33,7 @@ lora:
    - "v_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml
@ -38,6 +38,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
@ -37,6 +37,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_GLM-4.7-Flash.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_GLM-4.7-Flash.yaml
@ -35,6 +35,7 @@ lora:
    - "out_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml
@ -37,6 +37,7 @@ lora:
    - "out_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml
@ -33,6 +33,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml
@ -38,6 +38,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml
+++ b/studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml
@ -33,6 +33,7 @@ lora:
    - "v_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
+++ b/studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/imdatta0_tiny_qwen3_moe_2.8B_0.7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/imdatta0_tiny_qwen3_moe_2.8B_0.7B.yaml
@ -36,6 +36,7 @@ lora:
    - "gate_up_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml
@ -34,6 +34,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-30B-A3B-Instruct-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-30B-A3B-Instruct-2507.yaml
@ -36,6 +36,7 @@ lora:
    - "gate_up_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml
@ -35,6 +35,7 @@ lora:
    - "down_proj"
  use_rslora: false
  use_loftq: false
+  use_dora: false

 logging:
  enable_wandb: false
--- a/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml
+++ b/studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml
@ -29,6 +29,7 @@ lora:
    - "all-linear"
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/assets/configs/vision_lora.yaml
+++ b/studio/backend/assets/configs/vision_lora.yaml
@ -30,6 +30,7 @@ lora:
  vision_all_linear: true
  use_rslora: false
  use_loftq: false
+  use_dora: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
--- a/studio/backend/auth/terminal_prompt.py
+++ b/studio/backend/auth/terminal_prompt.py
@ -236,6 +236,10 @@ def prompt_for_password_change(
                out.write(f"Password must be at least {min_length} characters; try again.\n")
                out.flush()
                continue
+            if any(ch.isspace() for ch in new_password):
+                out.write("Password cannot contain spaces; try again.\n")
+                out.flush()
+                continue
            if is_current_password(new_password):
                out.write(
                    "New password must differ from the current bootstrap password; try again.\n"
--- a/studio/backend/cloudflare_tunnel.py
+++ b/studio/backend/cloudflare_tunnel.py
@ -20,6 +20,7 @@ import shutil
 import subprocess
 import sys
 import threading
+import time
 from pathlib import Path
 from typing import Optional, Tuple

@ -40,6 +41,22 @@ _RELEASE_BASE = "https://github.com/cloudflare/cloudflared/releases/latest/downl
 _READY_TIMEOUT = 15.0  # seconds to wait for the URL + a registered edge connection
 _DOWNLOAD_TIMEOUT = 60  # urlopen timeout for the one-time binary download

+# A registered edge connection does not mean the hostname resolves yet, so the
+# URL is fetched once before it is advertised.
+_PUBLIC_PROBE_PATH = "/api/health"
+_PUBLIC_PROBE_MARKER = "Unsloth UI Backend"
+# One deadline for DNS propagation + the health probe, bounding the startup stall.
+_PUBLIC_PROBE_TIMEOUT = 45.0
+_PUBLIC_PROBE_ATTEMPT_TIMEOUT = 5.0
+_PUBLIC_PROBE_RETRY_DELAY = 1.0
+
+# Wait for the hostname via DoH first: an early OS lookup negative-caches the
+# NXDOMAIN for up to 30 min.
+_DNS_POLL_DELAY = 2.0
+# Retry transient DoH failures, but give up fast when DoH is blocked outright.
+_DNS_MAX_DOH_ERRORS = 3
+_DOH_URL = "https://cloudflare-dns.com/dns-query?name={host}&type=A"
+

 def _windows_hidden_kwargs() -> dict:
    """Suppress a child console window on Windows; no-op elsewhere."""
@ -191,6 +208,59 @@ def ensure_cloudflared() -> Optional[str]:
        return None


+def _wait_for_dns(host: str, deadline: float) -> None:
+    import json
+    import urllib.request
+
+    errors = 0
+    while True:
+        answered = False
+        try:
+            req = urllib.request.Request(
+                _DOH_URL.format(host = host),
+                headers = {"Accept": "application/dns-json", "User-Agent": "unsloth-studio"},
+            )
+            with urllib.request.urlopen(req, timeout = 5) as response:
+                answered = bool(json.loads(response.read(65536)).get("Answer"))
+            errors = 0
+        except Exception:
+            errors += 1
+            if errors >= _DNS_MAX_DOH_ERRORS:
+                return
+        if answered:
+            return
+        remaining = deadline - time.monotonic()
+        if remaining <= 0:
+            return
+        time.sleep(min(_DNS_POLL_DELAY, remaining))
+
+
+def verify_public_url(url: str, timeout: float = _PUBLIC_PROBE_TIMEOUT) -> bool:
+    import json
+    import urllib.request
+    from urllib.parse import urlsplit
+
+    deadline = time.monotonic() + timeout
+    host = urlsplit(url).hostname
+    if host:
+        _wait_for_dns(host, deadline)
+
+    probe_url = f"{url.rstrip('/')}{_PUBLIC_PROBE_PATH}"
+    while True:
+        try:
+            req = urllib.request.Request(probe_url, headers = {"User-Agent": "unsloth-studio"})
+            with urllib.request.urlopen(req, timeout = _PUBLIC_PROBE_ATTEMPT_TIMEOUT) as response:
+                body = response.read(4096)
+            if json.loads(body).get("service") == _PUBLIC_PROBE_MARKER:
+                return True
+        except Exception:
+            pass
+        remaining = deadline - time.monotonic()
+        if remaining <= 0:
+            return False
+        time.sleep(min(_PUBLIC_PROBE_RETRY_DELAY, remaining))
+
+
 class CloudflareTunnel:
    """A cloudflared quick tunnel to http://localhost:<port>. Best-effort throughout.

@ -322,11 +392,12 @@ def start_studio_tunnel(port: int, timeout: float = _READY_TIMEOUT) -> Optional[
    """Start a quick tunnel and return its public URL once it is actually
    serving, or None (best-effort).

-    Waits for cloudflared to both mint the URL and register an edge connection
-    before returning, so the caller never advertises a URL that yields Cloudflare
-    error 1033 (HTTP 530). If a URL is minted but no connection registers within
-    the window (e.g. quic is blocked on this network), retries once forcing the
-    http2 protocol. On any failure the tunnel is stopped and None is returned.
+    Waits for cloudflared to both mint the URL and register an edge connection,
+    then fetches /api/health over the public URL, so the caller never advertises
+    a link that yields Cloudflare error 1033 (HTTP 530) or an unresolvable host.
+    If a URL is minted but no connection registers within the window (e.g. quic
+    is blocked on this network), retries once forcing the http2 protocol. On any
+    failure the tunnel is stopped and None is returned.
    """
    global _active_tunnel, _shutdown_requested
    binary = ensure_cloudflared()
@ -349,9 +420,13 @@ def start_studio_tunnel(port: int, timeout: float = _READY_TIMEOUT) -> Optional[
            prior, _active_tunnel = _active_tunnel, tunnel
        if prior is not None:
            prior.stop()
+        registered = False
        try:
            tunnel.start()
            url = tunnel.wait_for_ready(timeout)
+            registered = url is not None
+            if url and not verify_public_url(url):
+                url = None
        except Exception:
            url = None
        if url:
@ -371,6 +446,9 @@ def start_studio_tunnel(port: int, timeout: float = _READY_TIMEOUT) -> Optional[
        # http2 will not help, so do not burn another window on it.
        if not saw_url:
            return None
+        # probe failure after registering is DNS propagation; http2 would not help
+        if registered:
+            return None
    return None


--- a/studio/backend/colab.py
+++ b/studio/backend/colab.py
@ -1,9 +1,7 @@
 # SPDX-License-Identifier: AGPL-3.0-only
 # Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0

-"""
-Colab helpers for Unsloth Studio. Uses Colab's built-in proxy.
-"""
+"""Colab helpers for Unsloth Studio. Uses Colab's built-in proxy."""

 from pathlib import Path
 import sys
@ -22,11 +20,9 @@ logger = get_logger(__name__)


 def get_colab_url(port: int = 8888) -> str:
-    """
-    Get the Colab proxy URL for a port.
+    """Get the Colab proxy URL for a port.

-    Retries up to 3 times, validating the result is a real HTTPS Colab URL.
-    Falls back to http://localhost:{port} only when all attempts fail.
+    Retries 3x validating a real HTTPS Colab URL; falls back to localhost on failure.
    """
    import time as _time

@ -55,28 +51,243 @@ def get_colab_url(port: int = 8888) -> str:
    return fallback


-def show_link(port: int = 8888, *, _url: "str | None" = None):
-    """Display a styled clickable link to the UI.
-
-    *_url* is an optional pre-fetched proxy URL; pass it to avoid a second eval_js round-trip.
-    """
-    from IPython.display import display, HTML
-
-    url = _url if _url is not None else get_colab_url(port)
-
-    # Truncated display URL; try/except so an odd URL shape still renders the link.
+def _short_colab_url(url: str, port: int) -> str:
+    """Truncated display form of a Colab proxy URL; falls back to the full URL."""
    try:
        port_prefix = f"{port}-"
        idx = url.index(port_prefix)
        next_dash = url.index("-", idx + len(port_prefix))
-        short_url = url[: next_dash + 1] + "..."
+        return url[: next_dash + 1] + "..."
    except (ValueError, IndexError):
-        short_url = url
+        return url

-    # Plain-text line so the URL shows even if HTML display fails.
-    logger.info(f"🌐 Unsloth Studio URL: {url}")

-    html = f"""
+def _is_colab_proxy_url(url: str, port: int) -> bool:
+    """True when *url* looks like a real Colab kernel proxy, not a localhost fallback."""
+    return bool(url and isinstance(url, str) and url.startswith("https://") and str(port) in url)
+
+
+def _is_colab_runtime() -> bool:
+    """True on a hosted Colab notebook kernel.
+
+    Reuses the backend's main Colab detector (``/content`` + Colab env / ``google.colab``)
+    instead of a single env var, which is not always present on hosted runtimes.
+    """
+    try:
+        from main import _IS_COLAB
+        return bool(_IS_COLAB)
+    except Exception:
+        return False
+
+
+def _colab_login_credentials_path() -> Path:
+    from auth.storage import DB_PATH
+    return DB_PATH.parent / ".colab_notebook_login"
+
+
+def _store_colab_login_credentials(username: str, password: str) -> None:
+    """Persist Colab admin credentials for notebook re-runs after interrupt."""
+    path = _colab_login_credentials_path()
+    try:
+        path.parent.mkdir(parents = True, exist_ok = True)
+        path.write_text(f"{username}\n{password}\n")
+        try:
+            import os
+            os.chmod(path, 0o600)
+        except OSError:
+            pass
+    except OSError as e:
+        logger.info(f"Could not persist Colab login credentials ({e}).")
+
+
+def _load_colab_login_credentials() -> "tuple[str, str] | None":
+    """Return stored Colab admin credentials from a previous ``start()`` run, if any."""
+    path = _colab_login_credentials_path()
+    try:
+        if not path.is_file():
+            return None
+        lines = path.read_text().splitlines()
+        if len(lines) >= 2 and lines[0] and lines[1]:
+            return lines[0], lines[1]
+    except OSError as e:
+        logger.info(f"Could not load Colab login credentials ({e}).")
+    return None
+
+
+def _clear_colab_login_credentials() -> None:
+    """Drop the cached Colab credentials once they no longer authenticate."""
+    path = _colab_login_credentials_path()
+    try:
+        path.unlink(missing_ok = True)
+    except OSError as e:
+        logger.info(f"Could not clear Colab login credentials ({e}).")
+
+
+def _colab_credentials_still_valid(username: str, password: str) -> bool:
+    """True when *password* still matches the stored admin hash.
+
+    Guards against redisplaying a cached first-run password after the user has
+    changed the admin password through the app, which would print credentials
+    that no longer authenticate to the current Cloudflare tunnel.
+    """
+    try:
+        from auth.storage import get_user_and_secret
+        from auth.hashing import verify_password
+    except Exception as e:
+        logger.info(f"Could not load auth to validate cached Colab credentials ({e}).")
+        return False
+    try:
+        row = get_user_and_secret(username)
+        if not row:
+            return False
+        salt, pwd_hash = row[0], row[1]
+        return bool(verify_password(password, salt, pwd_hash))
+    except Exception as e:
+        logger.info(f"Could not validate cached Colab credentials ({e}).")
+        return False
+
+
+def _colab_wants_cloudflare(cloudflare: "bool | None") -> bool:
+    """Resolve whether to open a Cloudflare tunnel.
+
+    ``None`` auto-enables on real Colab (the in-cell proxy embed is often blank);
+    pass ``False`` to opt out.
+    """
+    if cloudflare is not None:
+        return cloudflare
+    return _is_colab_runtime()
+
+
+def _finalize_colab_admin_password() -> "tuple[str, str] | None":
+    """Clear the bootstrap-password gate on Colab so Cloudflare tunnels can start.
+
+    Returns ``(username, password)`` for display in the notebook. On first run the
+    random admin password is finalized; on later runs (e.g. after interrupt) the
+    stored credentials are re-displayed so the Cloudflare link stays usable.
+    Anyone who can read this cell already controls the runtime.
+    """
+    if not _is_colab_runtime():
+        return None
+    try:
+        from auth.storage import (
+            DEFAULT_ADMIN_USERNAME,
+            ensure_default_admin,
+            generate_bootstrap_password,
+            get_bootstrap_password,
+            requires_password_change,
+            update_password,
+        )
+    except Exception as e:
+        logger.warning(
+            f"Could not load auth for Colab setup ({e}); Cloudflare link may be blocked."
+        )
+        return None
+
+    try:
+        ensure_default_admin()
+        username = DEFAULT_ADMIN_USERNAME
+        if not requires_password_change(username):
+            creds = _load_colab_login_credentials()
+            if creds is not None and _colab_credentials_still_valid(username, creds[1]):
+                return creds
+            # The admin password was changed through the app after the first run,
+            # so the cached copy is stale; drop it instead of printing dead credentials.
+            _clear_colab_login_credentials()
+            return None
+        password = get_bootstrap_password() or generate_bootstrap_password()
+        if not update_password(username, password):
+            logger.warning(
+                "Could not finalize Colab admin password; Cloudflare link may be blocked."
+            )
+            return None
+        _store_colab_login_credentials(username, password)
+        return username, password
+    except Exception as e:
+        logger.warning(
+            f"Could not finalize Colab admin password ({e}); Cloudflare link may be blocked."
+        )
+        return None
+
+
+def _colab_login_html(username: str, password: str) -> str:
+    """Notebook card with Colab admin credentials (shown once after auto-finalize)."""
+    return f"""
+    <div style="display: inline-block; padding: 20px; background: #ffffff; border: 2px solid #000000;
+                border-radius: 12px; margin: 10px 0; font-family: system-ui, -apple-system, sans-serif;">
+        <h2 style="color: #000000; margin: 0 0 12px 0; font-size: 22px; font-weight: 800;">
+            Unsloth Studio Login (Colab)
+        </h2>
+        <p style="color: #333333; margin: 0 0 12px 0; font-size: 14px; font-weight: bold;">
+            Log in as <code>{username}</code> with this password. This cell is visible only in
+            your notebook session.
+        </p>
+        <p style="color: #333333; margin: 0; font-size: 14px; font-family: monospace; font-weight: bold;">
+            Password: <code>{password}</code>
+        </p>
+    </div>
+    """
+
+
+def _show_colab_login_credentials(username: str, password: str) -> None:
+    """Display Colab admin credentials in the notebook output."""
+    from IPython.display import HTML, display
+
+    logger.info(f"🔐 Unsloth Studio login — user: {username}")
+    display(HTML(_colab_login_html(username, password)))
+
+
+def _ready_card_html(
+    url: str,
+    port: int,
+    *,
+    has_cloudflare_link: bool = False,
+    cloudflare_requested: bool = False,
+) -> str:
+    """Branded ready card for the in-notebook Studio view.
+
+    Colab ``*.prod.colab.dev`` proxy URLs are session-scoped and 404 when opened as a
+    top-level tab or on another device, so never ``window.open`` them. On real Colab the
+    Cloudflare link is the supported entry point because in-cell proxy embeds often stay blank.
+    """
+    short_url = _short_colab_url(url, port)
+    if _is_colab_runtime() or _is_colab_proxy_url(url, port):
+        if has_cloudflare_link:
+            embed_note = (
+                "Open Studio with the Cloudflare link above. In-cell proxy previews on "
+                "current Colab often stay blank, so the tunnel link is the supported path."
+            )
+        elif cloudflare_requested:
+            embed_note = (
+                "Could not open a Cloudflare tunnel, so Studio may be unreachable on Colab. "
+                "Check the logs above and re-run this cell. Pass "
+                '<code style="background:#f3f3f3;padding:2px 6px;border-radius:4px;">'
+                "cloudflare=True</code> after fixing any tunnel errors."
+            )
+        else:
+            embed_note = (
+                "Colab proxy links cannot be opened in a new tab (they 404 outside this "
+                'notebook). Re-run with <code style="background:#f3f3f3;padding:2px 6px;'
+                'border-radius:4px;">start(cloudflare=True)</code> for a working link.'
+            )
+        return f"""
+    <div style="display: inline-block; padding: 20px; background: #ffffff; border: 2px solid #000000;
+                border-radius: 12px; margin: 10px 0; font-family: system-ui, -apple-system, sans-serif;">
+        <h2 style="color: #000000; margin: 0 0 12px 0; font-size: 26px; font-weight: 800;
+                   display: flex; align-items: center; gap: 12px;">
+            <img src="https://github.com/unslothai/unsloth/raw/main/studio/frontend/public/unsloth-gem.png"
+                 height="48" style="display:block;">
+            Unsloth Studio is Ready!
+        </h2>
+        <p style="color: #333333; margin: 0 0 8px 0; font-size: 15px; font-weight: bold;">
+            {embed_note}
+        </p>
+        <p style="color: #666666; margin: 16px 0 0 0; font-size: 13px; font-family: monospace; font-weight: bold;">
+            {short_url}
+        </p>
+    </div>
+    """
+
+    return f"""
    <div style="display: inline-block; padding: 20px; background: #ffffff; border: 2px solid #000000;
                border-radius: 12px; margin: 10px 0; font-family: system-ui, -apple-system, sans-serif;">
        <h2 style="color: #000000; margin: 0 0 12px 0; font-size: 26px; font-weight: 800;
@ -100,15 +311,52 @@ def show_link(port: int = 8888, *, _url: "str | None" = None):
        </p>
    </div>
    """
-    display(HTML(html))
+
+
+def show_link(
+    port: int = 8888,
+    *,
+    _url: "str | None" = None,
+    has_cloudflare_link: bool = False,
+    cloudflare_requested: bool = False,
+):
+    """Display a styled ready card for the UI.
+
+    Colab proxy URLs are informational only (no new-tab open; they 404 outside the cell);
+    non-proxy URLs keep a clickable open button. *_url* is an optional pre-fetched proxy
+    URL to avoid a second eval_js round-trip.
+    """
+    from IPython.display import display, HTML
+
+    url = _url if _url is not None else get_colab_url(port)
+    logger.info(f"🌐 Unsloth Studio URL: {url}")
+    display(
+        HTML(
+            _ready_card_html(
+                url,
+                port,
+                has_cloudflare_link = has_cloudflare_link,
+                cloudflare_requested = cloudflare_requested,
+            )
+        )
+    )
+
+
+def _warn_colab_cloudflare_missing(*, use_cloudflare: bool, cloudflare_url: "str | None") -> None:
+    """Log a prominent warning when Colab expected a tunnel but none was opened."""
+    if not use_cloudflare or cloudflare_url or not _is_colab_runtime():
+        return
+    logger.warning(
+        "Colab Cloudflare tunnel unavailable — Studio is unlikely to be reachable in this "
+        "notebook. Check the logs above for tunnel or auth errors, then re-run start()."
+    )


 def _bootstrap_password_pending() -> bool:
    """True while the default admin still owes a bootstrap-password change.

-    While pending, main.py injects that password into same-origin GETs, and a public
-    tunnel GET (no Origin) reads as same-origin, so sharing the link would leak admin
-    access. Fails safe to pending if the state cannot be read.
+    While pending, a public tunnel GET (no Origin) reads as same-origin and gets the
+    injected password, so sharing the link would leak admin access. Fails safe to pending.
    """
    try:
        from auth.storage import requires_password_change, DEFAULT_ADMIN_USERNAME
@ -121,9 +369,8 @@ def _bootstrap_password_pending() -> bool:
 def start_cloudflare_tunnel(port: int) -> "str | None":
    """Open a shareable Cloudflare quick tunnel to localhost:*port*, or None.

-    run_server suppresses the tunnel on Colab by design, so we start it directly.
-    Refused while the bootstrap password is pending; any failure collapses to None
-    and the Colab proxy still works.
+    run_server suppresses the tunnel on Colab, so we start it directly. Refused while the
+    bootstrap password is pending; any failure collapses to None (Colab proxy still works).
    """
    if _bootstrap_password_pending():
        logger.warning(
@ -152,9 +399,9 @@ def start_cloudflare_tunnel(port: int) -> "str | None":
 def _publish_cloudflare_url(cloudflare_url: "str | None") -> None:
    """Publish a directly-started tunnel URL onto app.state so /api/health advertises it.

-    run_server only sets this when it opens the tunnel itself, which it skips on Colab,
-    so we set it here. Otherwise the frontend's API examples fall back to an
-    unreachable server_url. Best-effort.
+    run_server sets this only when it opens the tunnel itself (skipped on Colab), so we
+    set it here; otherwise the frontend's API examples fall back to an unreachable
+    server_url. Best-effort.
    """
    if not cloudflare_url:
        return
@ -183,8 +430,7 @@ def _stop_cloudflare_tunnel() -> None:
 def _is_studio_healthy(port: int, timeout: float = 2.0) -> bool:
    """True only if Unsloth Studio (not some other app) answers /api/health on *port*.

-    The service-marker check stops the reuse path reusing or tunneling a foreign
-    process that merely serves /api/health.
+    The service-marker check stops the reuse path reusing or tunneling a foreign process.
    """
    import json, urllib.request
    try:
@ -194,8 +440,29 @@ def _is_studio_healthy(port: int, timeout: float = 2.0) -> bool:
        return False


-def _shareable_link_html(cloudflare_url: str) -> str:
-    """Branded card for the shareable Cloudflare link, styled like the show_link banner."""
+def _shareable_link_html(
+    cloudflare_url: str,
+    password: "str | None" = None,
+    username: "str | None" = None,
+) -> str:
+    """Branded card for the shareable Cloudflare link, styled like the show_link banner.
+
+    *password* renders under the link so the credential sits in the card with the button
+    it unlocks. The username is always the default admin, so it reads inline.
+    """
+    login_block = ""
+    if password:
+        login_block = f"""
+        <p style="color: #000000; margin: 16px 0 0 0; font-size: 20px; font-weight: 800;">
+            Password
+        </p>
+        <p style="margin: 6px 0 0 0;"><code style="display: inline-block; font-size: 24px;
+            font-weight: 800; text-decoration: underline; background: #f3f3f3;
+            padding: 4px 10px; border-radius: 6px;">{password}</code></p>
+        <p style="color: #666666; margin: 6px 0 0 0; font-size: 12px;">
+            Log in as <code>{username}</code> with this password. Shown only in your
+            notebook session, and never included in the shared link.
+        </p>"""
    return f"""
    <div style="display: inline-block; padding: 20px; background: #ffffff; border: 2px solid #000000;
                border-radius: 12px; margin: 10px 0; font-family: system-ui, -apple-system, sans-serif;">
@ -213,40 +480,55 @@ def _shareable_link_html(cloudflare_url: str) -> str:
            Open Unsloth Studio
        </a>
        <p style="color: #333333; margin: 12px 0 0 0; font-size: 14px; font-weight: bold;">
-            This Cloudflare HTTPS link works from any device — share it with anyone. The Colab view below only works in this tab.
+            This Cloudflare HTTPS link works from any device, so you can share it with anyone.
        </p>
        <p style="color: #333333; margin: 16px 0 0 0; font-size: 13px; font-family: monospace; font-weight: bold;">
-            🔗 {cloudflare_url}
-        </p>
+            🔗 <a href="{cloudflare_url}" onclick="var w=window.open(this.href,'_blank');if(!w){{return true;}}return false;"
+                  style="color: #000000; text-decoration: underline; cursor: pointer;">{cloudflare_url}</a>
+        </p>{login_block}
    </div>
    """


-def _show_and_embed(port: int, *, cloudflare_url: "str | None" = None):
-    """Render the Unsloth header + iframe for *port*, with a shareable-link card above
-    when *cloudflare_url* is set. Falls back to serve_kernel_port_as_iframe."""
-    url = get_colab_url(port)
-    logger.info(f"🌐 Unsloth Studio URL: {url}")
-    if cloudflare_url:
-        logger.info(f"🔗 Shareable Cloudflare link: {cloudflare_url}")
+# Height for serve_kernel_port_as_iframe (~82vh on a 1080p screen, clamped).
+_COLAB_IFRAME_HEIGHT = 900

+
+def _embed_kernel_port_iframe(port: int) -> bool:
+    """Embed Studio via Colab's native kernel-port iframe helper.
+
+    Only trusted on a real Colab runtime: colabtools can import ``google.colab`` and
+    queue browser-side JS without appending an iframe, so callers outside Colab must use
+    the HTML iframe path instead.
+    """
+    if not _is_colab_runtime():
+        return False
+    try:
+        from google.colab import output as colab_output
+    except ImportError:
+        return False
+    try:
+        colab_output.serve_kernel_port_as_iframe(
+            port,
+            height = _COLAB_IFRAME_HEIGHT,
+            width = "100%",
+        )
+        return True
+    except Exception as e:
+        logger.info(f"serve_kernel_port_as_iframe failed ({e}); trying HTML iframe.")
+        return False
+
+
+def _embed_html_iframe(url: str, port: int) -> bool:
+    """Fallback embed: raw HTML iframe when the Colab helper is unavailable."""
    try:
        from IPython.display import HTML, display
+    except ImportError:
+        return False

-        iframe_id = f"unsloth-studio-{port}"
-
-        # Truncated header URL — best-effort, falls back to full URL.
-        try:
-            port_prefix = f"{port}-"
-            idx = url.index(port_prefix)
-            next_dash = url.index("-", idx + len(port_prefix))
-            short_url = url[: next_dash + 1] + "..."
-        except (ValueError, IndexError):
-            short_url = url
-
-        if cloudflare_url:
-            display(HTML(_shareable_link_html(cloudflare_url)))
-
+    short_url = _short_colab_url(url, port)
+    iframe_id = f"unsloth-studio-{port}"
+    try:
        display(
            HTML(f"""
 <div style="font-family:system-ui,-apple-system,sans-serif;margin:8px 0;
@ -266,41 +548,110 @@ def _show_and_embed(port: int, *, cloudflare_url: "str | None" = None):
 </div>
 """)
        )
-    except Exception:
-        # Fallback: Colab's built-in helper.
+        return True
+    except Exception as e:
+        logger.info(f"HTML iframe embed failed ({e}).")
+        return False
+
+
+def _show_and_embed(
+    port: int,
+    *,
+    cloudflare_url: "str | None" = None,
+    colab_login: "tuple[str, str] | None" = None,
+    cloudflare_requested: bool = False,
+):
+    """Render the Unsloth ready card + iframe for *port*.
+
+    Prefer Colab's ``serve_kernel_port_as_iframe`` on real Colab; raw HTML iframe is the
+    fallback. Cloudflare cards stay clickable.
+    """
+    url = get_colab_url(port)
+    logger.info(f"🌐 Unsloth Studio URL: {url}")
+    if cloudflare_url:
+        logger.info(f"🔗 Shareable Cloudflare link: {cloudflare_url}")
+
+    _warn_colab_cloudflare_missing(
+        use_cloudflare = cloudflare_requested,
+        cloudflare_url = cloudflare_url,
+    )
+
+    # Fold the credentials into the link card rather than a second card below it.
+    credentials_shown = False
+    if cloudflare_url:
        try:
-            from google.colab import output as colab_output
-            colab_output.serve_kernel_port_as_iframe(port, height = 900, width = "100%")
-        except ImportError:
-            pass
+            from IPython.display import HTML, display
+
+            username, password = colab_login if colab_login else (None, None)
+            display(HTML(_shareable_link_html(cloudflare_url, password, username)))
+            credentials_shown = bool(colab_login)
+        except Exception as e:
+            logger.info(f"Could not render Cloudflare link card ({e}).")
+
+    if colab_login and not credentials_shown:
+        try:
+            _show_colab_login_credentials(*colab_login)
+        except Exception as e:
+            logger.info(f"Could not render Colab login card ({e}).")
+
+    # With a tunnel up the embed below is skipped, so the ready card would only restate
+    # the link card and print a proxy URL that 404s outside this tab.
+    skip_ready_card = _is_colab_runtime() and bool(cloudflare_url)
+    if not skip_ready_card:
+        try:
+            show_link(
+                port,
+                _url = url,
+                has_cloudflare_link = bool(cloudflare_url),
+                cloudflare_requested = cloudflare_requested,
+            )
+        except Exception as e:
+            logger.info(f"Could not render Unsloth link card ({e}).")
+
+    # On Colab with a working tunnel, skip the in-cell proxy embed (often blank).
+    if _is_colab_runtime() and cloudflare_url:
+        return
+
+    # Real Colab: kernel helper needs only the port (works when eval_js failed).
+    if _is_colab_runtime():
+        if _embed_kernel_port_iframe(port):
+            return
+    _embed_html_iframe(url, port)


-def start(port: int = 8888, *, cloudflare: bool = False):
+def start(port: int = 8888, *, cloudflare: "bool | None" = None):
    """Start Unsloth Studio in Colab and display the URL.

    Args:
        port: Port to bind/serve on.
-        cloudflare: Opt in to a shareable Cloudflare HTTPS link reachable from any
-            device (default OFF). It exposes Unsloth's login page beyond Colab, so it
-            stays an explicit opt-in; the default shows only the in-tab proxy iframe.
+        cloudflare: Shareable Cloudflare HTTPS link. ``None`` (default) auto-enables on
+            real Colab because the in-cell proxy embed is often blank; pass ``False`` to
+            skip the tunnel or ``True`` to force it on other runtimes.

    Usage:
-        start()                    # Colab-proxy iframe only (default)
-        start(cloudflare=True)     # also open a shareable Cloudflare link
+        start()                    # Cloudflare link on Colab (auto); proxy iframe elsewhere
+        start(cloudflare=False)    # Colab proxy iframe only (often blank on current Colab)
+        start(cloudflare=True)     # force Cloudflare link on any runtime
    """
    import time

    logger.info("🦥 Starting Unsloth Studio...")
+    use_cloudflare = _colab_wants_cloudflare(cloudflare)

-    # Fast path: Unsloth already running (cell re-run). Re-launching would collide on
-    # the port, so just re-show the link and iframe.
+    # Fast path: already running (cell re-run); re-show link/iframe instead of rebinding the port.
    if _is_studio_healthy(port):
        logger.info(f"   Unsloth is already running on port {port} — reusing existing server.")
        # try/finally: tear the tunnel down even if interrupted mid-start/render.
        try:
-            cf_url = start_cloudflare_tunnel(port) if cloudflare else None
+            colab_login = _finalize_colab_admin_password() if use_cloudflare else None
+            cf_url = start_cloudflare_tunnel(port) if use_cloudflare else None
            _publish_cloudflare_url(cf_url)
-            _show_and_embed(port, cloudflare_url = cf_url)
+            _show_and_embed(
+                port,
+                cloudflare_url = cf_url,
+                colab_login = colab_login,
+                cloudflare_requested = use_cloudflare,
+            )
            for _ in range(10000):
                time.sleep(300)
                print("=", end = "", flush = True)
@ -313,7 +664,6 @@ def start(port: int = 8888, *, cloudflare: bool = False):
    logger.info("   Loading backend...")
    from run import run_server

-    # Auto-detect frontend path
    repo_root = Path(__file__).parent.parent
    frontend_path = repo_root / "frontend" / "dist"

@ -323,8 +673,7 @@ def start(port: int = 8888, *, cloudflare: bool = False):

    logger.info("   Starting server...")
    try:
-        # cloudflare=False: this helper owns the tunnel (Colab's own
-        # start(cloudflare=...) drives it), so pin it off explicitly.
+        # cloudflare=False: this helper owns the tunnel (via start(cloudflare=...)), so pin it off.
        app = run_server(
            host = "0.0.0.0",
            port = port,
@ -339,14 +688,12 @@ def start(port: int = 8888, *, cloudflare: bool = False):
        logger.error(f"❌ Unsloth Studio failed to start: {exc}")
        return

-    # run_server auto-increments the port if in use; read back the bound port so the
-    # proxy URL and iframe point at the right place.
+    # run_server may auto-increment the port; read back the bound port for the proxy URL/iframe.
    actual_port: int = getattr(getattr(app, "state", None), "server_port", None) or port

    logger.info(f"   Server started on port {actual_port}!")

-    # Poll health endpoint before showing the link — avoids the race where ready_event
-    # fires but the process hasn't finished binding.
+    # Poll health before showing the link: avoids the race where ready_event fires pre-bind.
    import urllib.request

    server_ready = False
@ -365,12 +712,17 @@ def start(port: int = 8888, *, cloudflare: bool = False):
        )
        return

-    # Open the tunnel now the server is healthy, publish its URL for /api/health, and
-    # tear it down on interrupt (try/finally) rather than orphan the process.
+    # Server healthy: finalize Colab auth, open the tunnel, publish URL, tear down on interrupt.
    try:
-        cf_url = start_cloudflare_tunnel(actual_port) if cloudflare else None
+        colab_login = _finalize_colab_admin_password() if use_cloudflare else None
+        cf_url = start_cloudflare_tunnel(actual_port) if use_cloudflare else None
        _publish_cloudflare_url(cf_url)
-        _show_and_embed(actual_port, cloudflare_url = cf_url)
+        _show_and_embed(
+            actual_port,
+            cloudflare_url = cf_url,
+            colab_login = colab_login,
+            cloudflare_requested = use_cloudflare,
+        )

        # Keep kernel alive so the daemon server thread runs.
        for _ in range(10000):
--- a/studio/backend/core/data_recipe/jobs/manager.py
+++ b/studio/backend/core/data_recipe/jobs/manager.py
@ -27,7 +27,6 @@ from .constants import (
 )
 from .parse import apply_update, coerce_event, parse_log_message
 from .types import Job
-from .worker import run_job_process
 from loggers import get_logger

 logger = get_logger(__name__)
@ -169,12 +168,18 @@ class JobManager:
                native_path_secret_removed_for_child_start,
                run_without_native_path_secret,
            )
+            from utils.hf_cache_settings import child_environment_for_spawn, get_hf_cache_paths

-            with native_path_secret_removed_for_child_start():
+            cache_env = get_hf_cache_paths().child_env({})
+
+            with (
+                child_environment_for_spawn(cache_env),
+                native_path_secret_removed_for_child_start(),
+            ):
                mp_q = _CTX.Queue()
                proc = _CTX.Process(
                    target = run_without_native_path_secret,
-                    args = (run_job_process,),
+                    args = ("core.data_recipe.jobs.worker", "run_job_process", cache_env),
                    kwargs = {"event_queue": mp_q, "recipe": recipe, "run": run_payload},
                    daemon = True,
                )
--- a/studio/backend/core/export/export.py
+++ b/studio/backend/core/export/export.py
@ -81,6 +81,82 @@ _PYTORCH_MISSING_MESSAGE = (
 _LLAMA_CPP_SCRIPTS_WARNING_EMITTED = False


+def _multi_gpu_device_map_kwargs() -> dict:
+    """``device_map`` kwargs for sharding a checkpoint across every visible GPU.
+
+    unsloth's ``from_pretrained`` defaults to ``device_map="sequential"``, which stacks
+    the whole model on GPU0 and OOMs multi-GPU hosts whose other GPUs sit empty (#7053).
+    Returns ``{"device_map": "balanced"}`` only on a real multi-GPU CUDA/ROCm host
+    (mirroring the inference loader's ``get_device_map``), else empty so single-GPU, CPU
+    and MLX loads keep the loader default."""
+    if _IS_MLX:
+        return {}
+    try:
+        from utils.hardware import get_device_map, get_parent_visible_gpu_ids
+
+        visible = get_parent_visible_gpu_ids()
+        if len(visible) > 1:
+            device_map = get_device_map(visible)
+        elif not visible:
+            # UUID/MIG masks resolve to no numeric ids; get_device_map(None) falls back
+            # to the visible-GPU count, so a multi-GPU UUID/MIG host still shards.
+            device_map = get_device_map(None)
+        else:
+            return {}
+        if device_map == "balanced":
+            return {"device_map": device_map}
+    except Exception as exc:
+        logger.debug(f"multi-GPU device_map resolution failed; using loader default: {exc}")
+    return {}
+
+
+def _is_oom_error(exc: BaseException) -> bool:
+    """True for an accelerator OOM, however it is spelled.
+
+    accelerate and transformers re-raise it as a plain ``RuntimeError`` on several paths
+    and ROCm/XPU use their own classes, so match the message too.
+    """
+    if torch is not None:
+        oom_types = tuple(
+            t
+            for t in (
+                getattr(torch, "OutOfMemoryError", None),
+                getattr(getattr(torch, "cuda", None), "OutOfMemoryError", None),
+                getattr(getattr(torch, "xpu", None), "OutOfMemoryError", None),
+            )
+            if isinstance(t, type)
+        )
+        if oom_types and isinstance(exc, oom_types):
+            return True
+    return "out of memory" in f"{type(exc).__name__}: {exc}".lower()
+
+
+def _is_cpu_spill_rejection(exc: BaseException) -> bool:
+    """bitsandbytes refuses a map that spills to CPU/disk with a plain ``ValueError``.
+
+    Busy secondary GPUs can make ``balanced`` spill to CPU even where the old sequential
+    load fit on GPU0, and that message says nothing about memory, so the retry has to
+    match it explicitly. See transformers ``quantizers/quantizer_bnb_4bit.py``.
+    """
+    return "dispatched on the cpu or the disk" in str(exc).lower()
+
+
+class _CpuSpillRetry(Exception):
+    """A multi-GPU load that succeeded but left modules offloaded to CPU/disk."""
+
+
+def _cpu_offloaded_modules(model) -> int:
+    """Count the modules a load parked on CPU or disk.
+
+    Only bitsandbytes refuses such a map; a full-precision load accepts it, leaves the
+    parameters on meta and dies much later in safetensors with "Cannot copy out of meta
+    tensor". Nothing raises at load time, so inspect the map directly. PEFT re-dispatches
+    when attaching an adapter, so in practice this catches merged checkpoints.
+    """
+    device_map = getattr(model, "hf_device_map", None) or {}
+    return sum(1 for target in device_map.values() if str(target) in ("cpu", "disk"))
+
+
 def _supports_kwarg(fn, name):
    """True if `fn` accepts keyword `name` directly or via **kwargs."""
    import inspect
@ -271,6 +347,7 @@ class ExportBackend:
        load_in_4bit: bool = True,
        trust_remote_code: bool = False,
        hf_token: Optional[str] = None,
+        _device_map_override: Optional[dict] = None,
    ) -> Tuple[bool, str]:
        """
        Load a checkpoint for export.
@ -303,6 +380,14 @@ class ExportBackend:
            # Skip the Hub when offline so a no-internet export uses the local cache.
            local_files_only = _hf_offline()

+            # Shard across every visible GPU instead of stacking on GPU0 (#7053); {} on
+            # single-GPU/CPU/MLX. _device_map_override is the single-device retry below.
+            _device_map_kw = (
+                _multi_gpu_device_map_kwargs()
+                if _device_map_override is None
+                else _device_map_override
+            )
+
            # Run the type-detection probes in the forced-offline window (else a gated
            # base 404s); it covers is_vision_model's Hub reads + the transformers-5
            # subprocess, and local_files_only makes detect_audio_type's requests.get skip.
@ -328,6 +413,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

            elif self._audio_type == "whisper":
@ -343,6 +429,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

            elif self._audio_type == "snac":
@ -355,6 +442,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

            elif self._audio_type == "bicodec":
@ -368,6 +456,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

            elif self._audio_type == "dac":
@ -380,6 +469,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

            elif self.is_vision:
@ -392,6 +482,7 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )
                tokenizer = processor  # vision: processor acts as tokenizer

@ -405,8 +496,16 @@ class ExportBackend:
                    trust_remote_code = trust_remote_code,
                    token = token,
                    local_files_only = local_files_only,
+                    **_device_map_kw,
                )

+            # Only when we asked for the multi-GPU map: a single-GPU host has no second
+            # placement to retry on, so leave its behaviour untouched.
+            _offloaded = _cpu_offloaded_modules(model) if _device_map_kw else 0
+            if _device_map_override is None and _offloaded:
+                del model
+                raise _CpuSpillRetry(f"{_offloaded} module(s) offloaded to CPU/disk")
+
            if _IS_MLX:
                # MLX doesn't use PeftModel — detect LoRA via adapter_config.json
                self.is_peft = adapter_config.exists()
@ -429,11 +528,41 @@ class ExportBackend:
            return True, f"Loaded {model_type} model{peft_info} successfully"

        except Exception as e:
-            logger.error(f"Error loading checkpoint: {e}")
-            import traceback
+            # Sharding is an optimisation, never a requirement. "balanced" budgets from the
+            # free memory read BEFORE this process opens a CUDA context on each GPU, so when
+            # a training or chat job already owns the others the shard can OOM, or spill to
+            # CPU and be refused by bitsandbytes, where the old single-device load succeeded.
+            # Fall back once before giving up.
+            if (
+                _device_map_override is None
+                and (
+                    isinstance(e, _CpuSpillRetry) or _is_oom_error(e) or _is_cpu_spill_rejection(e)
+                )
+                and _multi_gpu_device_map_kwargs()
+            ):
+                # Retry outside this block: the live traceback pins the half-built model's
+                # frames, so an in-block retry inherits the exhausted device.
+                retry_reason = str(e)
+            else:
+                logger.error(f"Error loading checkpoint: {e}")
+                import traceback

-            logger.error(traceback.format_exc())
-            return False, f"Failed to load checkpoint: {str(e)}"
+                logger.error(traceback.format_exc())
+                return False, f"Failed to load checkpoint: {str(e)}"
+
+        logger.warning(
+            f"Multi-GPU export load unusable ({retry_reason}); retrying on "
+            f"the single-device loader default."
+        )
+        self.cleanup_memory()
+        return self.load_checkpoint(
+            checkpoint_path,
+            max_seq_length = max_seq_length,
+            load_in_4bit = load_in_4bit,
+            trust_remote_code = trust_remote_code,
+            hf_token = hf_token,
+            _device_map_override = {},
+        )

    def _write_export_metadata(self, save_directory: str):
        """Write export_metadata.json with base model info for Chat page discovery."""
@ -1048,6 +1177,21 @@ class ExportBackend:
                    "Use the safetensors adapter instead.",
                    None,
                )
+            # llama.cpp's convert_lora_to_gguf.py has no concept of DoRA's
+            # lora_magnitude_vector tensors: it only reads the standard
+            # lora_A/lora_B delta, so exporting a DoRA adapter would silently
+            # drop the magnitude rescaling and produce a GGUF LoRA file that
+            # loads fine but no longer matches the trained model.
+            _peft_config = getattr(self.current_model, "peft_config", {}).get("default")
+            if getattr(_peft_config, "use_dora", False):
+                return (
+                    False,
+                    "GGUF LoRA export is not supported for DoRA adapters: the GGUF LoRA "
+                    "format has no way to represent DoRA's magnitude vectors, so the "
+                    "exported file would silently lose the DoRA behavior. Use the "
+                    "safetensors adapter instead, or merge to a full GGUF model.",
+                    None,
+                )
            outtype = str(gguf_outtype).lower()
            if outtype not in _GGUF_LORA_OUTTYPES:
                return (
--- a/studio/backend/core/export/orchestrator.py
+++ b/studio/backend/core/export/orchestrator.py
@ -230,16 +230,20 @@ class ExportOrchestrator:
            native_path_secret_removed_for_child_start,
            run_without_native_path_secret,
        )
+        from utils.hf_cache_settings import child_environment_for_spawn, get_hf_cache_paths

-        from .worker import run_export_process
+        cache_env = get_hf_cache_paths().child_env({})

-        with native_path_secret_removed_for_child_start():
+        with (
+            child_environment_for_spawn(cache_env),
+            native_path_secret_removed_for_child_start(),
+        ):
            self._cmd_queue = _CTX.Queue()
            self._resp_queue = _CTX.Queue()

            self._proc = _CTX.Process(
                target = run_without_native_path_secret,
-                args = (run_export_process,),
+                args = ("core.export.worker", "run_export_process", cache_env),
                kwargs = {
                    "cmd_queue": self._cmd_queue,
                    "resp_queue": self._resp_queue,
--- a/studio/backend/core/inference/audio_codecs.py
+++ b/studio/backend/core/inference/audio_codecs.py
@ -76,8 +76,14 @@ class AudioCodecManager:
        if self._snac_model is not None:
            return
        from snac import SNAC
+        from utils.hf_cache_settings import active_hf_hub_cache

-        self._snac_model = SNAC.from_pretrained("hubertsiuzdak/snac_24khz").to(device).eval()
+        # Route weights to the selected cache; this can run in the main process.
+        self._snac_model = (
+            SNAC.from_pretrained("hubertsiuzdak/snac_24khz", cache_dir = active_hf_hub_cache())
+            .to(device)
+            .eval()
+        )
        logger.info("Loaded SNAC codec (24kHz)")

    def _load_bicodec(
--- a/studio/backend/core/inference/inference.py
+++ b/studio/backend/core/inference/inference.py
@ -8,6 +8,7 @@ from unsloth.chat_templates import get_chat_template
 from transformers import TextIteratorStreamer, TextStreamer
 from peft import PeftModel, PeftModelForCausalLM

+import contextlib
 import json
 import sys
 import torch
@ -1942,8 +1943,30 @@ class InferenceBackend:
            + text
            + "<|text_end|>\n<|audio_start|><|global_features_start|>\n"
        )
+
        with torch.inference_mode():
-            with torch.amp.autocast("cuda", dtype = model.dtype):
+            # Derive the autocast device from the loaded model, not from the
+            # global backend: a CPU-fallback DAC on an XPU/CUDA host must not
+            # open a GPU autocast context around CPU tensors.
+            device_type = (
+                model.device.type
+                if hasattr(model.device, "type")
+                else str(model.device).split(":", 1)[0]
+            )
+            # Clamp to autocast-supported backends so exotic devices
+            # (e.g. "meta" during accelerate offloaded loading) do not raise.
+            # MPS is autocast-supported since torch 2.3, keep it in the set.
+            if device_type not in ("cuda", "xpu", "mps", "cpu"):
+                device_type = "cpu"
+            # CPU and XPU autocast only accept bfloat16/float16. For a
+            # float32 model, skip autocast entirely to avoid raising or
+            # producing a warning on every generate call.
+            autocast_dtype_supported = model.dtype in (torch.bfloat16, torch.float16)
+            if device_type in ("cpu", "xpu") and not autocast_dtype_supported:
+                autocast_ctx = contextlib.nullcontext()
+            else:
+                autocast_ctx = torch.amp.autocast(device_type, dtype = model.dtype)
+            with autocast_ctx:
                inputs = tokenizer([prompt], return_tensors = "pt").to(model.device)
                generated = model.generate(
                    **inputs,
--- a/Show more
+++ b/Show more