From 1cc77061ab19d0874a3d51fa0c69e110f6644294 Mon Sep 17 00:00:00 2001 From: Roland Tannous Date: Mon, 6 Apr 2026 16:25:07 +0000 Subject: [PATCH] split venv_t5 into venv_t5_530 and venv_t5_550 for tiered transformers 5.x support --- studio/backend/core/export/worker.py | 3 - studio/backend/core/inference/worker.py | 2 - studio/backend/core/training/worker.py | 17 +++++ studio/setup.ps1 | 83 +++++++++++++++++++++++++ studio/setup.sh | 23 +++++++ 5 files changed, 123 insertions(+), 5 deletions(-) diff --git a/studio/backend/core/export/worker.py b/studio/backend/core/export/worker.py index f77b1966c4..1be456d4d4 100644 --- a/studio/backend/core/export/worker.py +++ b/studio/backend/core/export/worker.py @@ -181,11 +181,8 @@ def _setup_log_capture(resp_queue: Any) -> None: def _activate_transformers_version(model_name: str) -> None: """Activate the correct transformers version BEFORE any ML imports.""" - # Ensure backend is on path for utils imports - backend_path = str(Path(__file__).resolve().parent.parent.parent) if backend_path not in sys.path: sys.path.insert(0, backend_path) - from utils.transformers_version import activate_transformers_for_subprocess activate_transformers_for_subprocess(model_name) diff --git a/studio/backend/core/inference/worker.py b/studio/backend/core/inference/worker.py index fbcce276ba..4f86f99ded 100644 --- a/studio/backend/core/inference/worker.py +++ b/studio/backend/core/inference/worker.py @@ -35,8 +35,6 @@ from utils.hardware import apply_gpu_ids def _activate_transformers_version(model_name: str) -> None: """Activate the correct transformers version BEFORE any ML imports.""" - # Ensure backend is on path for utils imports - backend_path = str(Path(__file__).resolve().parent.parent.parent) if backend_path not in sys.path: sys.path.insert(0, backend_path) diff --git a/studio/backend/core/training/worker.py b/studio/backend/core/training/worker.py index 8ab2b5b2be..0b0a73313f 100644 --- a/studio/backend/core/training/worker.py +++ b/studio/backend/core/training/worker.py @@ -374,6 +374,7 @@ def run_training_process( ) return +<<<<<<< HEAD # ── 1a. Auto-enable trust_remote_code for NemotronH/Nano models ── # NemotronH has config parsing bugs in transformers that require # trust_remote_code=True as a workaround. Other transformers 5.x models @@ -385,6 +386,22 @@ def run_training_process( if ( any(sub in _lowered for sub in _NEMOTRON_TRUST_SUBSTRINGS) and (_lowered.startswith("unsloth/") or _lowered.startswith("nvidia/")) +======= + # ── 1a. Auto-enable trust_remote_code for unsloth/* transformers 5.x models ── + # Some newer architectures (e.g. NemotronH) have config parsing bugs in + # transformers that require trust_remote_code=True as a workaround. + # Only auto-enable for unsloth/* prefixed models (trusted source). + # Exclude Gemma 4 since it is a native transformers 5.5 model and + # trust_remote_code=True would bypass the compiler (disabling fused CE). + from utils.transformers_version import get_transformers_tier + + _lowered = model_name.lower() + _tier = get_transformers_tier(model_name) + if ( + _tier != "default" + and _lowered.startswith("unsloth/") + and _tier != "550" # Gemma 4 is native t5.5 — trust_remote_code bypasses compiler +>>>>>>> 970219a3 (split venv_t5 into venv_t5_530 and venv_t5_550 for tiered transformers 5.x support) and not config.get("trust_remote_code", False) ): config["trust_remote_code"] = True diff --git a/studio/setup.ps1 b/studio/setup.ps1 index 24ef3ef1eb..802ec1f21d 100644 --- a/studio/setup.ps1 +++ b/studio/setup.ps1 @@ -1701,6 +1701,89 @@ if ($stackExit -ne 0) { exit 1 } +# ── Pre-install transformers 5.x into .venv_t5_530/ and .venv_t5_550/ ── +# Models like GLM-4.7-Flash, Qwen3 MoE need transformers>=5.3.0. +# Gemma 4 models need transformers>=5.5.0. +# Pre-install into separate directories to avoid runtime pip overhead. +# The training subprocess prepends the appropriate dir to sys.path. +Write-Host "" + +# Clean up legacy single .venv_t5 directory +$VenvT5Legacy = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5" +if (Test-Path $VenvT5Legacy) { Remove-Item -Recurse -Force $VenvT5Legacy } + +$prevEAP_t5 = $ErrorActionPreference +$ErrorActionPreference = "Continue" + +# --- .venv_t5_530 (transformers 5.3.0) --- +substep "pre-installing transformers 5.3.0 for newer model support..." +$VenvT5_530Dir = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5_530" +if (Test-Path $VenvT5_530Dir) { Remove-Item -Recurse -Force $VenvT5_530Dir } +New-Item -ItemType Directory -Path $VenvT5_530Dir -Force | Out-Null +foreach ($pkg in @("transformers==5.3.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) { + if ($script:UnslothVerbose) { + Fast-Install --target $VenvT5_530Dir --no-deps $pkg + $t5PkgExit = $LASTEXITCODE + $output = "" + } else { + $output = Fast-Install --target $VenvT5_530Dir --no-deps $pkg | Out-String + $t5PkgExit = $LASTEXITCODE + } + if ($t5PkgExit -ne 0) { + Write-Host "[FAIL] Could not install $pkg into .venv_t5_530/" -ForegroundColor Red + Write-Host $output -ForegroundColor Red + $ErrorActionPreference = $prevEAP_t5 + exit 1 + } +} +if ($script:UnslothVerbose) { + Fast-Install --target $VenvT5_530Dir tiktoken + $tiktokenInstallExit = $LASTEXITCODE + $output = "" +} else { + $output = Fast-Install --target $VenvT5_530Dir tiktoken | Out-String + $tiktokenInstallExit = $LASTEXITCODE +} +if ($tiktokenInstallExit -ne 0) { + substep "Could not install tiktoken into .venv_t5_530/ -- Qwen tokenizers may fail" "Yellow" +} +step "transformers" "5.3.0 pre-installed" + +# --- .venv_t5_550 (transformers 5.5.0) --- +substep "pre-installing transformers 5.5.0 for Gemma 4 support..." +$VenvT5_550Dir = Join-Path $env:USERPROFILE ".unsloth\studio\.venv_t5_550" +if (Test-Path $VenvT5_550Dir) { Remove-Item -Recurse -Force $VenvT5_550Dir } +New-Item -ItemType Directory -Path $VenvT5_550Dir -Force | Out-Null +foreach ($pkg in @("transformers==5.5.0", "huggingface_hub==1.8.0", "hf_xet==1.4.2")) { + if ($script:UnslothVerbose) { + Fast-Install --target $VenvT5_550Dir --no-deps $pkg + $t5PkgExit = $LASTEXITCODE + $output = "" + } else { + $output = Fast-Install --target $VenvT5_550Dir --no-deps $pkg | Out-String + $t5PkgExit = $LASTEXITCODE + } + if ($t5PkgExit -ne 0) { + Write-Host "[FAIL] Could not install $pkg into .venv_t5_550/" -ForegroundColor Red + Write-Host $output -ForegroundColor Red + $ErrorActionPreference = $prevEAP_t5 + exit 1 + } +} +if ($script:UnslothVerbose) { + Fast-Install --target $VenvT5_550Dir tiktoken + $tiktokenInstallExit = $LASTEXITCODE + $output = "" +} else { + $output = Fast-Install --target $VenvT5_550Dir tiktoken | Out-String + $tiktokenInstallExit = $LASTEXITCODE +} +if ($tiktokenInstallExit -ne 0) { + substep "Could not install tiktoken into .venv_t5_550/ -- Qwen tokenizers may fail" "Yellow" +} +$ErrorActionPreference = $prevEAP_t5 +step "transformers" "5.5.0 pre-installed" + } else { step "python" "dependencies up to date" # Restore ErrorActionPreference (was lowered for pip/python section) diff --git a/studio/setup.sh b/studio/setup.sh index 9f21af534c..2b2c14e9b4 100755 --- a/studio/setup.sh +++ b/studio/setup.sh @@ -548,6 +548,29 @@ fi if [ "$_SKIP_PYTHON_DEPS" = false ]; then install_python_stack + + # ── 6b. Pre-install transformers 5.x into .venv_t5_530/ and .venv_t5_550/ ── + # Models like GLM-4.7-Flash, Qwen3 MoE need transformers>=5.3.0. + # Gemma 4 models need transformers>=5.5.0. + # Pre-install into separate directories to avoid runtime pip overhead. + # The training subprocess prepends the appropriate dir to sys.path. + + # Clean up legacy single .venv_t5 directory + [ -d "$STUDIO_HOME/.venv_t5" ] && rm -rf "$STUDIO_HOME/.venv_t5" + + mkdir -p "$VENV_T5_530_DIR" + run_quiet "install transformers 5.3.0" fast_install --target "$VENV_T5_530_DIR" --no-deps "transformers==5.3.0" + run_quiet "install huggingface_hub for t5_530" fast_install --target "$VENV_T5_530_DIR" --no-deps "huggingface_hub==1.8.0" + run_quiet "install hf_xet for t5_530" fast_install --target "$VENV_T5_530_DIR" --no-deps "hf_xet==1.4.2" + run_quiet "install tiktoken for t5_530" fast_install --target "$VENV_T5_530_DIR" "tiktoken" + step "transformers" "5.3.0 pre-installed" + + mkdir -p "$VENV_T5_550_DIR" + run_quiet "install transformers 5.5.0" fast_install --target "$VENV_T5_550_DIR" --no-deps "transformers==5.5.0" + run_quiet "install huggingface_hub for t5_550" fast_install --target "$VENV_T5_550_DIR" --no-deps "huggingface_hub==1.8.0" + run_quiet "install hf_xet for t5_550" fast_install --target "$VENV_T5_550_DIR" --no-deps "hf_xet==1.4.2" + run_quiet "install tiktoken for t5_550" fast_install --target "$VENV_T5_550_DIR" "tiktoken" + step "transformers" "5.5.0 pre-installed" else step "python" "dependencies up to date" verbose_substep "python deps check: installed=$_PKG_NAME@${INSTALLED_VER:-unknown} latest=${LATEST_VER:-unknown}"