Commit graph

4,859 commits

Author SHA1 Message Date
Alkın Ünlü
6180adda1b fix(trainer): import psutil to prevent NameError in _prepare_dataset (#3780)
* fix(trainer): import psutil to prevent NameError in _prepare_dataset

Fixes #3777

* Update rl.py

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
2025-12-28 21:18:02 -08:00
Daniel Han
f40fa7a0e8 Update FUNDING.yml (#3792) 2025-12-28 19:57:43 -08:00
Michael Han
96de7a817d Update README for new unsloth.ai/docs.md 2025-12-27 00:49:19 -08:00
Fizza Mukhtar
f57cd25d46 Clarify NotImplementedError for fast_inference with full_finetuning (#3768)
* Improve error message for fast_inference and full_finetuning

* Refine error message string formatting

* Update unsloth/models/vision.py

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
2025-12-25 18:46:13 -08:00
Strahinja Stamenkovic
a058885b8a Add missing import of inspect (#3778)
* Add missing import of inspect

* Update device_type.py
2025-12-25 18:43:59 -08:00
pre-commit-ci[bot]
7c1d528c00 [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
2025-12-23 20:06:37 +00:00
numb3r33
3aaca08a0e Refactor return statement replacement to use explicit newlines
Replace f-string triple-quoted approach with explicit newline characters
for clearer string construction in the grpo_trainer patch.
2025-12-24 01:34:21 +05:30
pre-commit-ci[bot]
d8c9e6aafb [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
2025-12-24 01:34:21 +05:30
numb3r33
fe21809bda Fix indentation handling in grpo_trainer return statement replacement
Use regex to dynamically detect and preserve the original indentation
when replacing the 'return output' statement, instead of hardcoding
spaces. This ensures the patched code maintains consistent indentation
regardless of the original formatting.
2025-12-24 01:34:21 +05:30
pre-commit-ci[bot]
c8e7bd9f09 [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
2025-12-24 01:34:21 +05:30
numb3r33
a6405b36b4 Remove the comment. 2025-12-24 01:34:21 +05:30
abhishek.sharma
91671433b0 Fix model training state restoration in GRPO trainer
Store the model's training state before generation and restore inference
mode after completion if the model wasn't originally in training mode.
This ensures the model returns to the correct state after generate and
score operations.
2025-12-24 01:34:21 +05:30
Daniel Han
dea670a1b6 Merge branch 'main' into nightly 2025-12-23 05:51:04 -08:00
Daniel Han
1ff6fc85f0 llama.cpp fixes 2025-12-23 05:50:26 -08:00
Daniel Han
cbfa7a20f9 Update rl.py 2025-12-23 05:42:58 -08:00
Daniel Han
0ae7d2ba28 Update rl.py 2025-12-23 05:35:06 -08:00
Daniel Han
a4408f0e50 Merge branch 'main' into nightly 2025-12-23 04:52:58 -08:00
Daniel Han
fd42103a9b Nightly (#3767)
* Update _utils.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [FIX] [Transformers] VLM input embeds fix for gradients (#3715)

* Fix get_input_embeds call for VLMs

* patch input_require_grads instead

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old patch

* cleanup old patch

* cleanup

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* use logger instead of prints

* Move unsloth present set

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Update rope_embedding.py

* Fixes

* Update _utils.py

* Update import_fixes.py

* Update rl_replacements.py

* fix_openenv_no_vllm

* Fix

* Update __init__.py

* Update __init__.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* logger

* Update __init__.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update __init__.py

* Update import_fixes.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update import_fixes.py

* Update unsloth/import_fixes.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update save.py

* [fbgemm] Silence tma fbgemm (#3735)

* Silence fbgemm TMA print

Also safer .push_to_hub

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Update loader.py

* Update save.py

* Update save.py

* Update _utils.py

* Update _utils.py

* Diffusers warnings

* Update pyproject.toml

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [hf_hub] Token login (#3739)

* login on token

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old code

* safer imports

* cleanup

* Return token after login

* correct return types

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* add back imports

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* finish return token

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Do not overwrite slots (#3752)

* Do not overwrite slots

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Update save.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2025-12-23 04:52:29 -08:00
pre-commit-ci[bot]
ad73b7e493 [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
2025-12-23 12:51:09 +00:00
Daniel Han
6866c4b3df Update save.py 2025-12-23 04:46:43 -08:00
Daniel Han
10ef983541 Merge branch 'main' into nightly 2025-12-23 04:46:15 -08:00
Daniel Han
691b5d129f Update save.py 2025-12-23 00:55:06 -08:00
pre-commit-ci[bot]
e134ceed79 [pre-commit.ci] pre-commit autoupdate (#3760)
updates:
- [github.com/astral-sh/ruff-pre-commit: v0.14.9 → v0.14.10](https://github.com/astral-sh/ruff-pre-commit/compare/v0.14.9...v0.14.10)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2025-12-22 23:12:00 -08:00
Daniel Han
ad5b2f66af Merge branch 'main' into nightly 2025-12-19 19:37:52 -08:00
Daniel Han
25a3141663 Update loader.py 2025-12-19 19:37:49 -08:00
Daniel Han
47ede31b8c Nightly (#3753)
* Update _utils.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [FIX] [Transformers] VLM input embeds fix for gradients (#3715)

* Fix get_input_embeds call for VLMs

* patch input_require_grads instead

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old patch

* cleanup old patch

* cleanup

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* use logger instead of prints

* Move unsloth present set

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Update rope_embedding.py

* Fixes

* Update _utils.py

* Update import_fixes.py

* Update rl_replacements.py

* fix_openenv_no_vllm

* Fix

* Update __init__.py

* Update __init__.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* logger

* Update __init__.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update __init__.py

* Update import_fixes.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update import_fixes.py

* Update unsloth/import_fixes.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update save.py

* [fbgemm] Silence tma fbgemm (#3735)

* Silence fbgemm TMA print

Also safer .push_to_hub

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Update loader.py

* Update save.py

* Update save.py

* Update _utils.py

* Update _utils.py

* Diffusers warnings

* Update pyproject.toml

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [hf_hub] Token login (#3739)

* login on token

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old code

* safer imports

* cleanup

* Return token after login

* correct return types

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* add back imports

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* finish return token

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Do not overwrite slots (#3752)

* Do not overwrite slots

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2025-12-19 19:35:41 -08:00
Daniel Han
bfa7301768 Merge branch 'main' into nightly 2025-12-19 19:31:23 -08:00
Daniel Han
9a6d703d3c Update _utils.py 2025-12-19 19:24:49 -08:00
Strahinja Stamenkovic
490153500b Enable 4-bit quantization on AMD Radeon GPUs (#3748)
* Enable 4-bit quant on Radeon

* Fix table centering

* Update comments for clarity

* Handle failure to import Bitsandbytes

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update device_type.py

* Apply suggestion from @danielhanchen

* Update device_type.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
2025-12-19 19:22:56 -08:00
Dan Saunders
1a507b4a82 Fix VLM DDP checkpointing (#3751) 2025-12-19 19:09:16 -08:00
Datta Nimmaturi
3918e07df8 Do not overwrite slots (#3752)
* Do not overwrite slots

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2025-12-19 19:08:28 -08:00
Daniel Han
3f238efba5 Merge branch 'main' into nightly 2025-12-18 09:30:19 -08:00
Daniel Han
a36eb9b9a1 FunctionGemma 2025-12-18 09:27:46 -08:00
DoubleMathew
96bd2a7668 Fix Deepseek OCR Lora Model Load (#3738)
* fix deepseek ocr lora_model load: trust_remote_code option

check for import error in autoconfig/peftconfig from_pretrained error

handle import

* Apply suggestion from @gemini-code-assist[bot]

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Apply suggestion from @gemini-code-assist[bot]

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Apply suggestion from @gemini-code-assist[bot]

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Apply suggestion from @gemini-code-assist[bot]

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

---------

Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2025-12-18 04:09:17 -08:00
Datta Nimmaturi
6832ce8098 [hf_hub] Token login (#3739)
* login on token

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old code

* safer imports

* cleanup

* Return token after login

* correct return types

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* add back imports

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* finish return token

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
2025-12-18 04:07:12 -08:00
Dan Saunders
44104fa83d review comments 2025-12-17 15:28:03 -05:00
Dan Saunders
7833191626 review comments 2025-12-17 15:22:42 -05:00
Daniel Han
6b676abaad Merge branch 'main' into nightly 2025-12-17 03:32:57 -08:00
Daniel Han
1e7302cd77 Nightly (#3737)
* Update _utils.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [FIX] [Transformers] VLM input embeds fix for gradients (#3715)

* Fix get_input_embeds call for VLMs

* patch input_require_grads instead

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* cleanup old patch

* cleanup old patch

* cleanup

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Apply suggestion from @danielhanchen

* use logger instead of prints

* Move unsloth present set

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Daniel Han <danielhanchen@gmail.com>

* Update rope_embedding.py

* Fixes

* Update _utils.py

* Update import_fixes.py

* Update rl_replacements.py

* fix_openenv_no_vllm

* Fix

* Update __init__.py

* Update __init__.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* logger

* Update __init__.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update __init__.py

* Update import_fixes.py

* Update __init__.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* Update import_fixes.py

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Update import_fixes.py

* Update unsloth/import_fixes.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update save.py

* [fbgemm] Silence tma fbgemm (#3735)

* Silence fbgemm TMA print

Also safer .push_to_hub

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Update loader.py

* Update save.py

* Update save.py

* Update _utils.py

* Update _utils.py

* Diffusers warnings

* Update pyproject.toml

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Datta Nimmaturi <venkatadattasainimmaturi@gmail.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2025-12-17 03:31:48 -08:00
pre-commit-ci[bot]
ec0b96012e [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
2025-12-17 11:29:40 +00:00
Daniel Han
150f7e017e Update pyproject.toml 2025-12-17 03:26:19 -08:00
Daniel Han
37888f3e10 Diffusers warnings 2025-12-17 03:25:40 -08:00
Daniel Han
fecee1b386 Update _utils.py 2025-12-17 02:54:15 -08:00
Daniel Han
2f7132d7c9 Update _utils.py 2025-12-17 02:38:05 -08:00
Daniel Han
be4b5996d7 Update save.py 2025-12-17 02:28:03 -08:00
Daniel Han
7d82555dc8 Update save.py 2025-12-17 02:21:47 -08:00
Daniel Han
28120b8a88 Update loader.py 2025-12-17 01:51:29 -08:00
Daniel Han
b342b6cf28 Merge branch 'nightly' of https://github.com/unslothai/unsloth into nightly 2025-12-17 01:07:29 -08:00
Datta Nimmaturi
a32f28d30b [fbgemm] Silence tma fbgemm (#3735)
* Silence fbgemm TMA print

Also safer .push_to_hub

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
2025-12-17 01:07:21 -08:00
Daniel Han
59ed4fce4c Update save.py 2025-12-16 23:15:31 -08:00