ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge remote-tracking branch 'origin/main' into windows-torch-211

Browse source 
Reconcile PR #7256 (Windows torch 2.11 + release preservation, installer
comment reduction, torch 2.11 default line) with 77 commits of main install
rewrites. Base is main's newer install semantics; this branch's still-novel
contributions are layered on top.

Key decisions:
- install.ps1: kept main's rollback lifecycle (try/finally Restore-StudioVenvRollback,
  #7342) and the installed-version report (#7265); layered this branch's torch-2.11
  allowance (torch<2.12.0 on the Windows CUDA fresh-install, CPU fallback and flavor
  repairs), the release-preservation port (Get-InstalledTorchVersionRaw / kept-release
  installs / UNSLOTH_KEPT_TORCH handoff / torch-overrides freeze) and the Exit-InstallFailure
  UNSLOTH_KEPT_TORCH clear.
- install.sh: took main's newer AMD/Strix routing (per-arch index reroute #7264/#7300,
  runtime-less gfx inference #7305, KFD detection fix #7314, Radeon 8065S regex #7290,
  signal-restore trap #7342) and the #7365 unsloth pin bump; kept this branch's
  _TORCH_CEILING/_TORCHVISION_CEILING/_TORCHAUDIO_CEILING refactor widening the default
  ceiling to torch<2.12.0.
- studio/install_python_stack.py: deferred all five code conflicts to main's Strix
  inference logic (result is AST-identical to main; only comment reductions remain).
- studio/setup.ps1: took main's comment covering the whisper.cpp dictation markers
  (#7095); the UNSLOTH_KEPT_TORCH consumption handoff auto-merged intact.

Dropped as superseded by main: this branch's stale unsloth>=2026.7.4 pins (main #7365),
the older rocm7.1->rocm7.2 Strix reroute (main #7264/#7300), the narrow Radeon 80[0-9]0S
regex (main #7290), and the pre-inference has_hip_torch gate (main's rocm_torch_ready gate,
Codex P1 #7305). The two review-item fixes (grep -E in test_torch_constraint.sh,
UNSLOTH_KEPT_TORCH clear in Exit-InstallFailure) survive.
This commit is contained in:
danielhanchen 2026-07-24 11:24:55 +00:00
commit a14cc540ae
588 changed files with 58517 additions and 11322 deletions
Download patch file Download diff file Expand all files Collapse all files

33
.github/workflows/cross-platform-parity-ci.yml vendored
View file

@ -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

13
.github/workflows/studio-backend-ci.yml vendored
View file

@ -193,6 +193,7 @@ 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' \
@ -205,14 +206,15 @@ 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
@ -228,6 +230,7 @@ jobs:
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_staged_validation_enabled.sh \
tests/sh/test_tauri_install_exit_order.sh \
tests/sh/test_torch_constraint.sh \
tests/sh/test_torch_flavor.sh \

60
.github/workflows/studio-ui-smoke.yml vendored
View file

@ -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

7
README.md
View file

@ -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.

125
install.ps1
View file

@ -1295,13 +1295,82 @@ 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
@ -1313,7 +1382,9 @@ 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"
@ -1328,11 +1399,13 @@ 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
}
}
# Raw torch.__version__ from $PythonExe's venv (last non-empty stdout line), or $null.
@ -1366,6 +1439,8 @@ exit 0
} catch { return $null }
}
$studioVenvReplacementCommitted = $false
try {
if (Test-Path -LiteralPath $VenvPython) {
# env-mode: $StudioHome is a user-chosen workspace, so refuse to nuke an existing venv lacking Unsloth sentinels (-PathType Leaf rejects a directory at the sentinel path; accept the in-VENV ownership marker so partial-install retries aren't blocked).
if (
@ -1711,7 +1786,7 @@ exit 0
$nameArchTable = @(
@{ P = "9070 XT|9080"; A = "gfx1201" } # RDNA 4 (RX 9070 XT / 9080)
@{ P = "9070|9060"; A = "gfx1200" } # RDNA 4 (RX 9070 / 9060)
@{ P = "8060S|8050S|8040S|Strix Halo|Ryzen AI Max|AI Max"; A = "gfx1151" } # RDNA 3.5 (Strix Halo: Radeon 8060S/8050S/8040S iGPU, Ryzen AI Max+)
@{ P = "8065S|8060S|8050S|8040S|Strix Halo|Ryzen AI Max|AI Max"; A = "gfx1151" } # RDNA 3.5 (Strix Halo + Gorgon Halo: Radeon 8065S/8060S/8050S/8040S iGPU, Ryzen AI Max / Max+)
@{ P = "890M|880M|860M|840M|Strix Point|Krackan|HX 37[05]|AI 9 HX|AI 9 36[05]|AI 7 35[05]|AI 5 34[05]|AI 7 PRO 35|AI 5 33"; A = "gfx1150" } # RDNA 3.5 (Strix/Krackan Point: Radeon 890M/880M iGPU, Ryzen AI 9 HX 370/375)
@{ P = "RX 7900|RX 7800|RX 7700(?!S)|PRO W7900|PRO W7800|PRO W7700"; A = "gfx1100" } # RDNA 3 desktop/workstation (Navi 31)
@{ P = "RX 7600|RX 7700S|RX 7650|PRO W7600|PRO W7500|PRO V710"; A = "gfx1102" } # RDNA 3 (Navi 33)
@ -1880,16 +1955,18 @@ exit 0
# Strip userinfo AND query/fragment so an authenticated pin never leaks. Shared with _strip_index_url_credentials (install.sh / py / setup.ps1).
function Remove-IndexUrlCredentials {
param([string]$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}://${host_}"
@ -2006,6 +2083,10 @@ exit 0
"gfx1151" = "gfx1151"; "gfx1150" = "gfx1150" # RDNA 3.5 (Strix Halo/Point)
"gfx1103" = "gfx110X-all"; "gfx1102" = "gfx110X-all" # RDNA 3
"gfx1101" = "gfx110X-all"; "gfx1100" = "gfx110X-all"
"gfx1036" = "gfx103X-all"; "gfx1035" = "gfx103X-all" # RDNA 2 (RX 6000)
"gfx1034" = "gfx103X-all"; "gfx1033" = "gfx103X-all"
"gfx1032" = "gfx103X-all"; "gfx1031" = "gfx103X-all"
"gfx1030" = "gfx103X-all"
"gfx90a" = "gfx90a"; "gfx908" = "gfx908" # MI200/MI100
}
# gfx120X (RDNA 4) and gfx1151/gfx1150 (Strix) hit a null-pointer bug in torch._C._grouped_mm on torch <2.11.0 (TheRock #5284 / #3284); force torch>=2.11.0 so pip skips the broken 2.10.0 wheels on the AMD index. The <2.12.0 ceiling (matches install_python_stack.py) blocks an unvalidated future 2.12.0+rocm wheel; bump both when 2.12.x is confirmed on gfx120X / Strix.
@ -2139,7 +2220,7 @@ exit 0
substep "upgrading unsloth in migrated environment..."
if ($SkipTorch) {
# No-torch: install unsloth + unsloth-zoo --no-deps, then runtime deps (typer, safetensors, transformers, etc.) --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 the matching pydantic-core (no-torch-runtime.txt below is --no-deps); all transitive deps are torch-free.
$baseInstallExit = Invoke-InstallCommandRetry -Label "install pydantic" { uv pip install --python $VenvPython pydantic }
@ -2151,7 +2232,7 @@ 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 $baseInstallExit)" -ForegroundColor Red
@ -2258,7 +2339,7 @@ exit 0
substep "installing unsloth (this may take a few minutes)..."
if ($SkipTorch) {
# No-torch: install unsloth + unsloth-zoo --no-deps, then runtime deps (typer, safetensors, transformers, etc.) --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 }
@ -2273,11 +2354,11 @@ exit 0
# Freeze the installed torch trio so this with-deps resolve can't downgrade the pinned +cuXXX/+rocm build (twin of install.sh's _build_unsloth_torch_overrides).
$script:TorchOverridesFile = New-UnslothTorchOverridesFile -PythonExe $VenvPython
if ($script:TorchOverridesFile) {
$baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (local)" { uv pip install --python $VenvPython --upgrade-package unsloth --overrides $script:TorchOverridesFile "unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4" }
$baseInstallExit = Invoke-InstallCommandRetry -Label "install unsloth (local)" { uv pip install --python $VenvPython --upgrade-package unsloth --overrides $script:TorchOverridesFile "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
Remove-Item -LiteralPath $script:TorchOverridesFile -Force -ErrorAction SilentlyContinue
$script:TorchOverridesFile = $null
} else {
$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 $VenvPython --upgrade-package unsloth "unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6" }
}
} else {
# Freeze the installed torch trio (see above) so the with-deps unsloth resolve can't strip the +cuXXX/+rocm suffix.
@ -2314,7 +2395,7 @@ exit 0
Write-TauriLog "STEP" "Installing unsloth"
substep "installing unsloth (this may take a few minutes)..."
if ($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)
@ -2340,6 +2421,13 @@ 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+cpu against a CUDA index and setup.ps1 loops on "cpu != required cuXXX". Reinstall the right triplet when a GPU build is expected: CUDA from $TorchIndexUrl, ROCm from $ROCmIndexUrl (a PEP 503 index uv resolves via --default-index). --no-torch / CPU-only hosts are no-ops. ──
if (-not $SkipTorch) {
$expectedTorchTag = Get-ExpectedTorchFlavorTag -TorchIndexUrl $TorchIndexUrl -ROCmIndexUrl $ROCmIndexUrl
@ -2601,6 +2689,13 @@ 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 $env:Path) would win.
if ($StudioRedirectMode -eq 'env' -and (Test-Path -LiteralPath $ShimExe)) {

554
install.sh
View file

@ -444,14 +444,20 @@ _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" "$_candidate"; then
_VENV_ROLLBACK_ACTIVE=false
_VENV_ROLLBACK_DIR=""
return 1
fi
substep "previous environment preserved for rollback"
}
@ -472,13 +478,68 @@ _restore_studio_venv_replacement() {
fi
}
_commit_studio_venv_replacement() {
[ "$_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 "$_UNSLOTH_TORCH_OVERRIDES" 2>/dev/null || true
}
_on_install_exit() {
@ -486,14 +547,28 @@ _on_install_exit() {
if [ "$_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"
}
# Empty so an inherited value never reaches the trap's rm.
_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() {
@ -519,6 +594,36 @@ _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' "$_ad_label"
)
}
# ── Helper: install packages via apt, escalating to sudo only if needed ──
# Usage: _smart_apt_install pkg1 pkg2 pkg3 ...
_smart_apt_install() {
@ -549,11 +654,14 @@ _smart_apt_install() {
# Step 3: Escalate -- need elevated permissions for remaining packages
if command -v sudo >/dev/null 2>&1; then
_ad_desc="$(_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] "
@ -1541,8 +1649,8 @@ _maybe_reroute_strixhalo_to_2404() {
[ -e /dev/dxg ] || return 0
# A usable NVIDIA GPU means the CUDA path works here, so don't reroute for AMD.
if _has_usable_nvidia_gpu; then return 0; fi
# Strix APUs show in /proc/cpuinfo; discrete cards don't, so also try WMI.
if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9]0S|Strix Halo' /proc/cpuinfo 2>/dev/null \
# 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][05]S|Strix Halo' /proc/cpuinfo 2>/dev/null \
&& ! _wsl_amd_gpu_name >/dev/null 2>&1; then
return 0
fi
@ -1949,17 +2057,152 @@ _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/ && $2 == 4098 { found = 1 } END { exit !found }' \
/sys/class/kfd/kfd/topology/nodes/*/properties 2>/dev/null; then
# vendor_id 4098 = 0x1002 (AMD); the NVIDIA open kernel module can register KFD nodes too.
# vendor_id 4098 = 0x1002 (AMD) marks a GPU node: the KFD CPU node
# reports vendor_id 0, so any 4098 node is an AMD GPU. NVIDIA's open
# kernel module (driver 560+) registers KFD nodes as vendor_id 4318
# (0x10DE), so this never false-positives on NVIDIA-only hosts.
# The prior check also required a gpu_id line, but gpu_id is a SIBLING
# sysfs file, not a line in properties -- it never matched, so the
# fallback silently missed every ROCm-less AMD host (issue: fresh
# Arch/CachyOS boxes reporting "no GPU detected").
return 0
fi
return 1
}
# ── Detect GPU and choose PyTorch index URL (mirrors Get-TorchIndexUrl) ──
# CPU-only machines get the cpu index, avoiding the --torch-backend=auto solver dead-end.
# Returns 0 if an AMD display GPU is on the PCI bus even when ROCm can't use it
# (e.g. a Strix Halo iGPU with no /dev/kfd). Only sharpens the "no GPU detected"
# hint. vendor 0x1002 = AMD/ATI; class 0x03* = display controller.
_amd_gpu_present_via_pci() {
[ -d /sys/bus/pci/devices ] || return 1
for _pci_vendor in /sys/bus/pci/devices/*/vendor; do
[ -r "$_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 ;;
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 "$1" in
*"9070 XT"*|*9080*) echo gfx1201 ;;
*9070*|*9060*) echo gfx1200 ;;
*"8065S"*|*"8060S"*|*"8050S"*|*"8040S"*|*"Strix Halo"*|*"Ryzen AI Max"*|*"AI Max"*) echo gfx1151 ;;
*"890M"*|*"880M"*|*"860M"*|*"840M"*|*"Strix Point"*|*"Krackan"*|*"HX 37"*|*"AI 9 HX"*|*"AI 9 36"*|*"AI 7 35"*|*"AI 5 34"*|*"AI 7 PRO 35"*|*"AI 5 33"*) echo gfx1150 ;;
*"RX 7600"*|*"RX 7700S"*|*"RX 7650"*|*"PRO W7600"*|*"PRO W7500"*|*"PRO V710"*) echo gfx1102 ;;
*"RX 7900"*|*"RX 7800"*|*"RX 7700"*|*"PRO W7900"*|*"PRO W7800"*|*"PRO W7700"*) echo gfx1100 ;;
*"780M"*|*"760M"*|*"740M"*|*"Phoenix"*|*"Hawk Point"*|*"Z1 Extreme"*|*"Z2 Extreme"*) echo gfx1103 ;;
*"RX 6900"*|*"RX 6800"*|*"RX 6750"*|*"RX 6700"*|*"PRO W6800"*|*"PRO W6900"*) echo gfx1030 ;;
*"RX 6650"*|*"RX 6600"*|*"PRO W6600"*|*"PRO W6650"*) echo gfx1032 ;;
*"RX 6500"*|*"RX 6400"*|*"RX 6300"*|*"PRO W6400"*|*"PRO W6500"*) echo gfx1034 ;;
*) return 1 ;;
esac
}
# Best-effort gfx inference when ROCm tools can't see the GPU (unslothai#7301).
# Mirrors install.ps1 arch resolution on Windows ($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 "$_gpu_evidence" ] && grep -qiE '890M|880M|860M|840M|Strix Point|Krackan|HX 37[05]|AI 9 HX|AI 9 36[05]|AI 7 35[05]|AI 5 34[05]|AI 7 PRO 35|AI 5 33' /proc/cpuinfo 2>/dev/null; then
echo gfx1150
return 0
fi
if command -v lspci >/dev/null 2>&1; then
# A non-AMD controller can enumerate first (Intel/ASPEED before an AMD
# dGPU), so scan every display-class line and take the first AMD one
# that maps. The vendor guard is case-SENSITIVE (a -i "ATI" would match
# "CorporATIon" on every Intel/NVIDIA line); whole-line matching also
# survives the 0000: PCI domain prefix. Mirrors install_python_stack.py.
_amd_disp=$(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
# dead-end where --torch-backend=auto resolves to unsloth==2024.8.
get_torch_index_url() {
_base="${UNSLOTH_PYTORCH_MIRROR:-https://download.pytorch.org/whl}"
_base="${_base%/}"
@ -2000,6 +2243,29 @@ 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 && \
@ -2016,7 +2282,11 @@ 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)
@ -2049,10 +2319,27 @@ get_torch_index_url() {
esac
return
fi
# ROCm version unreadable from any source: warn rather than silently install 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 "$_base/cpu"; return
fi
# Parse CUDA version from nvidia-smi (POSIX-safe): accept both "CUDA Version:" and the newer "CUDA UMD Version:". Bounded, C locale.
@ -2387,8 +2674,9 @@ _maybe_bootstrap_rocm_wsl() {
fi
# WSL GPU passthrough device must exist (present on any WSL2 GPU host).
[ -e /dev/dxg ] || return 0
# Strix APUs show in /proc/cpuinfo; discrete cards don't, so also ask the Windows host.
if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9]0S|Strix Halo' /proc/cpuinfo 2>/dev/null \
# Strix APUs show in /proc/cpuinfo (the CPU model); discrete cards don't, so also
# ask the Windows host. Either signal suffices; the bootstrap detects arch from rocminfo.
if ! grep -qiE 'Ryzen AI Max|Radeon 80[0-9][05]S|Strix Halo' /proc/cpuinfo 2>/dev/null \
&& ! _wsl_amd_gpu_name >/dev/null 2>&1; then
return 0
fi
@ -2463,7 +2751,84 @@ fi
TORCH_INDEX_URL=$(get_torch_index_url)
# Export UNSLOTH_TORCH_BACKEND ("cuda"/"rocm"/"cpu") for downstream scripts; classify on the FINAL lowercased leaf only so a custom mirror path containing "rocm"/"gfx" can't mislabel a cu*/cpu index.
# 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)
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.
# Classify on the FINAL path segment only: a custom UNSLOTH_PYTORCH_MIRROR
# whose base path happens to contain "rocm" or "gfx" must not mislabel a
# cu*/cpu index as ROCm (radeon repo URLs end in rocm-rel-X.Y/, Strix
# overrides in gfxNNNN/, so the trailing slash is stripped first).
# Lowercase the leaf so every gfx*/rocm*/cu* arm matches regardless of case (canonical AMD
# RDNA4 leaf is gfx120X-all). CUDA is branded only on a real cu[0-9]* leaf, so a mirror
# leaf (/current) does NOT commit a CUDA backend; an unknown leaf leaves the var unset so
# the stack probes the GPU. Query/fragment dropped first, then ALL trailing slashes (in
# lockstep with the shared _torch_index_url_leaf extractor).
_torch_index_leaf="${TORCH_INDEX_URL%%\?*}"
_torch_index_leaf="${_torch_index_leaf%%#*}"
# Strip ALL trailing slashes: .../cu128// must yield cu128, not empty.
@ -2508,20 +2873,64 @@ case "$TORCH_INDEX_URL" in
fi
;;
esac
# ── Strix Halo / Strix Point: force rocm7.2 wheels, bypass Radeon repo ───────
# gfx1151/gfx1150 segfault in torch._grouped_mm on ROCm 7.1.
case "$TORCH_INDEX_URL" in
*/rocm7.1|*/rocm7.1.*)
# Index gfx tokens by HIP_VISIBLE_DEVICES / ROCR_VISIBLE_DEVICES so a mixed Strix iGPU + selected dGPU box isn't rerouted.
_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}')
# 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.
# || 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}')
# Mirror the PowerShell `amd-smi static --asic` probe as a fallback.
_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=$( (unset ROCR_VISIBLE_DEVICES HIP_VISIBLE_DEVICES; amd-smi static --asic 2>/dev/null) | grep -oE 'gfx[1-9][0-9a-z]{2,3}' || true)
fi
fi
_runtime_gfx=""
@ -2546,13 +2955,14 @@ case "$TORCH_INDEX_URL" in
case "$_runtime_gfx" in
gfx1151|gfx1150) _strix_gfx="$_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] $_strix_gfx (Strix) detected -- routing to the AMD arch-specific index" >&2
echo " [WARN] torch 2.11+rocm7.13 has AMD's real gfx1150/gfx1151 fixes (the ROCm 7.1" >&2
echo " [WARN] _grouped_mm segfault, moe_utils.py:167, and later Strix kernel bugs)," >&2
echo " [WARN] and is more reliable than the rocm7.2 index or an offline Radeon repo." >&2
echo "" >&2
# AMD's arch-specific index has the real _grouped_mm fix (torch 2.11.0+rocm7.13.0); UNSLOTH_AMD_ROCM_MIRROR overrides for air-gapped installs.
_amd_strix_base="${UNSLOTH_AMD_ROCM_MIRROR:-https://repo.amd.com/rocm/whl}"
@ -2629,7 +3039,7 @@ elif case "$TORCH_INDEX_URL" in */rocm*|*/gfx*) true ;; *) false ;; esac; then
case "$_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+)
*"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"*|*"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)
@ -2664,6 +3074,17 @@ 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
@ -2672,9 +3093,20 @@ 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 [ "$OS" = "wsl" ]; then
# WSL + no GPU: often an AMD GPU whose ROCm-on-WSL runtime isn't exposed yet.
if [ "$_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" ] && [ "$_torch_index_pinned" = false ]; then
# WSL + no GPU detected (detection above found nothing). Common
# cause: an AMD GPU whose ROCm-on-WSL runtime isn't exposed yet --
# /dev/dxg present (graphics) but no ROCm runtime.
_wsl_ubu_ver=""
[ -r /etc/os-release ] && _wsl_ubu_ver=$(. /etc/os-release 2>/dev/null; printf '%s' "${VERSION_ID:-}")
if [ -e /dev/dxg ]; then
@ -2698,6 +3130,13 @@ case "$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," "$C_WARN"
substep " rocminfo, or amd-smi). Install the ROCm kernel stack so /dev/kfd exists;"
substep " Strix Halo (gfx1151/gfx1150) needs a recent kernel (6.11+) and ROCm 7.x."
fi
fi
substep "Re-run with --no-torch for GGUF-only (faster, no PyTorch):"
substep " curl -fsSL https://unsloth.ai/install.sh | sh -s -- --no-torch"
@ -2747,7 +3186,7 @@ if [ "$_MIGRATED" = true ]; then
# No-torch: --no-deps installs (PyPI metadata still hard-deps torch), then torch-free runtime deps --no-deps.
run_install_cmd_retry "install unsloth (migrated no-torch)" uv pip install --python "$_VENV_PY" --no-deps \
--reinstall-package unsloth --reinstall-package unsloth-zoo \
"unsloth>=2026.7.4" "unsloth-zoo>=2026.7.4"
"unsloth>=2026.7.5" "unsloth-zoo>=2026.7.6"
# Resolve pydantic WITH deps so its pydantic-core matches (all torch-free).
run_install_cmd_retry "install pydantic (with deps for compatible core)" \
uv pip install --python "$_VENV_PY" pydantic
@ -2761,7 +3200,7 @@ 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 "$_UNSLOTH_TORCH_OVERRIDES"
_UNSLOTH_TORCH_OVERRIDES=""
fi
@ -2949,7 +3388,7 @@ elif [ -n "$TORCH_INDEX_URL" ]; then
# No-torch: install unsloth + unsloth-zoo --no-deps, then runtime deps --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
@ -2968,7 +3407,7 @@ 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..."
@ -2995,7 +3434,7 @@ 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 "$_REPO_ROOT" --no-deps
substep "overlaying unsloth-zoo from git main..."
@ -3007,6 +3446,15 @@ else
fi
fi
_installed_package_version=$("$_VENV_PY" -c \
'from importlib.metadata import version; import sys; print(version(sys.argv[1]))' \
"$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, so uv keeps a stale torch==X+cpu against a GPU index; reinstall the right triplet, else warn loudly.
if [ "$SKIP_TORCH" = false ] && [ -n "${TORCH_INDEX_URL:-}" ]; then

32
pyproject.toml
View file

@ -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]",
]
@ -532,16 +543,19 @@ 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]",
]
kaggle = [
"unsloth[huggingface]",
@ -580,7 +594,7 @@ 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",
@ -831,16 +845,19 @@ 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]",
]
flashattentiontorch260abiFALSEcu12x = [
"flash-attn @ https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.6cxx11abiFALSE-cp39-cp39-linux_x86_64.whl ; ('linux' in sys_platform) and python_version == '3.9'",
@ -1125,7 +1142,8 @@ 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]",
@ -1279,6 +1297,7 @@ 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]",
@ -1297,6 +1316,7 @@ 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]

71
scripts/build_whisper_cpp.sh Executable file
View file

@ -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"

4
scripts/notebook_validator.py
View file

@ -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"},

94
scripts/scan_packages_baseline.json
View file

@ -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"
}
]
}

273
studio/Unsloth_Studio_Colab.ipynb
View file

@ -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
}

1
studio/backend/assets/configs/full_finetune.yaml
View file

@ -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

1
studio/backend/assets/configs/lora_text.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/default.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/embedding/unsloth_Qwen3-Embedding-0.6B.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/embedding/unsloth_all-MiniLM-L6-v2.yaml
View file

@ -30,6 +30,7 @@ lora:
- "query"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/embedding/unsloth_bge-m3.yaml
View file

@ -30,6 +30,7 @@ lora:
- "value"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/embedding/unsloth_embeddinggemma-300m.yaml
View file

@ -33,6 +33,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/embedding/unsloth_gte-modernbert-base.yaml
View file

@ -29,6 +29,7 @@ lora:
- "Wqkv"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-21B-A3B-PT.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/ernie/unsloth_ERNIE-4.5-VL-28B-A3B-PT.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/falcon/tiiuae_Falcon-H1-0.5B-Instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_codegemma-7b-bnb-4bit.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_functiongemma-270m-it.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-27b-bnb-4bit.yaml
View file

@ -33,6 +33,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-2-2b.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-270m-it.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-27b-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3-4b-pt.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-3n-E4B.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-26B-A4B.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-31B.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E2B.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B-it.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gemma/unsloth_gemma-4-E4B.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-120b.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/gpt-oss/unsloth_gpt-oss-20b.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-350m-unsloth-bnb-4bit.yaml
View file

@ -37,6 +37,7 @@ lora:
- "shared_mlp.output_linear"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/granite/unsloth_granite-4.0-h-micro.yaml
View file

@ -37,6 +37,7 @@ lora:
- "shared_mlp.output_linear"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-11B-Vision-Instruct.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-1B-Instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.2-3B-Instruct.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Llama-3.3-70B-Instruct.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-70B-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_Meta-Llama-3.1-8B-Instruct-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-Instruct-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llama/unsloth_llama-3-8b-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/llasa/unsloth_Llasa-3B.yaml
View file

@ -30,6 +30,7 @@ lora:
- "v_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_Magistral-Small-2509-unsloth-bnb-4bit.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_Ministral-3-3B-Instruct-2512.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Nemo-Base-2407-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_Mistral-Small-Instruct-2409.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_Pixtral-12B-2409.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-instruct-v0.3-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/mistral/unsloth_mistral-7b-v0.3-bnb-4bit.yaml
View file

@ -33,6 +33,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/OuteAI_Llama-OuteTTS-1.0-1B.yaml
View file

@ -33,6 +33,7 @@ lora:
- "v_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/Spark-TTS-0.5B_LLM.yaml
View file

@ -38,6 +38,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/sesame_csm-1b.yaml
View file

@ -37,6 +37,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_GLM-4.7-Flash.yaml
View file

@ -35,6 +35,7 @@ lora:
- "out_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_LFM2-1.2B.yaml
View file

@ -29,6 +29,7 @@ lora:
- "all-linear"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_Nemotron-3-Nano-30B-A3B.yaml
View file

@ -37,6 +37,7 @@ lora:
- "out_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_PaddleOCR-VL.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/other/unsloth_answerdotai_ModernBERT-large.yaml
View file

@ -33,6 +33,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_orpheus-3b-0.1-ft.yaml
View file

@ -38,6 +38,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_tinyllama-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/other/unsloth_whisper-large-v3.yaml
View file

@ -33,6 +33,7 @@ lora:
- "v_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3-medium-4k-instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-3.5-mini-instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/phi/unsloth_Phi-4.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/imdatta0_tiny_qwen3_moe_2.8B_0.7B.yaml
View file

@ -36,6 +36,7 @@ lora:
- "gate_up_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-7B.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2-VL-7B-Instruct.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-1.5B-Instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-7B.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-1.5B-Instruct.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-14B-Instruct.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-Coder-7B-Instruct-bnb-4bit.yaml
View file

@ -34,6 +34,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen2.5-VL-7B-Instruct-bnb-4bit.yaml
View file

@ -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

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-0.6B.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B-Base-unsloth-bnb-4bit.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-14B.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-30B-A3B-Instruct-2507.yaml
View file

@ -36,6 +36,7 @@ lora:
- "gate_up_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-32B.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Instruct-2507.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-4B-Thinking-2507.yaml
View file

@ -35,6 +35,7 @@ lora:
- "down_proj"
use_rslora: false
use_loftq: false
use_dora: false
logging:
enable_wandb: false

1
studio/backend/assets/configs/model_defaults/qwen/unsloth_Qwen3-VL-8B-Instruct-unsloth-bnb-4bit.yaml
View file

@ -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

1
studio/backend/assets/configs/vision_lora.yaml
View file

@ -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

4
studio/backend/auth/terminal_prompt.py
View file

@ -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"

88
studio/backend/cloudflare_tunnel.py
View file

@ -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

486
studio/backend/colab.py
View file

@ -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 +51,244 @@ 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 to Studio with the Cloudflare link above using these credentials. 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;">
Username: <code>{username}</code><br>
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 +312,52 @@ 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 +370,8 @@ 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 +400,9 @@ 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 +431,7 @@ 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:
@ -222,31 +469,45 @@ def _shareable_link_html(cloudflare_url: str) -> str:
"""
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 +527,101 @@ 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,
)
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
display(HTML(_shareable_link_html(cloudflare_url)))
except Exception as e:
logger.info(f"Could not render Cloudflare link card ({e}).")
if colab_login:
try:
_show_colab_login_credentials(*colab_login)
except Exception as e:
logger.info(f"Could not render Colab login card ({e}).")
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 +634,6 @@ 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 +643,7 @@ 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 +658,12 @@ 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 +682,17 @@ 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):

11
studio/backend/core/data_recipe/jobs/manager.py
View file

@ -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,
)

15
studio/backend/core/export/export.py
View file

@ -1048,6 +1048,21 @@ 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 (

10
studio/backend/core/export/orchestrator.py
View file

@ -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,

8
studio/backend/core/inference/audio_codecs.py
View file

@ -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(

25
studio/backend/core/inference/inference.py
View file

@ -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,

Some files were not shown because too many files have changed in this diff Show more