Daniel Han
0a00c6fee1
Update rl_replacements.py
2025-07-17 06:44:13 -07:00
Daniel Han
585453b19b
Update rl_replacements.py
2025-07-17 06:39:59 -07:00
Daniel Han
a14ee1b764
Update rl.py
2025-07-17 06:30:26 -07:00
Daniel Han
1a969535e4
Update rl_replacements.py
2025-07-17 06:26:10 -07:00
Daniel Han
34c66f0327
Update rl_replacements.py
2025-07-17 06:25:21 -07:00
Daniel Han
31df063c28
Update rl_replacements.py
2025-07-17 06:24:33 -07:00
Daniel Han
e4e6bc1e4a
Update rl_replacements.py
2025-07-17 06:23:37 -07:00
Daniel Han
bf628efd28
Update pyproject.toml
2025-07-17 05:38:04 -07:00
Daniel Han
fa7b673a9c
Update pyproject.toml
2025-07-17 05:26:58 -07:00
Daniel Han
743010d7fc
Merge branch 'main' into nightly
2025-07-17 05:14:52 -07:00
Daniel Han
170fbcad66
Revert "GRPO Fix - Support vllm pre-dequantized quantization states in fast_dequantize kernel ( #2943 )"
...
This reverts commit 2d4ef50fb3
2025-07-17 05:02:08 -07:00
Daniel Han
935754dba2
Update _utils.py
2025-07-17 02:08:49 -07:00
Daniel Han
d17519d819
compiler stance
2025-07-17 01:40:49 -07:00
Daniel Han
54f09f2fd7
Update vision.py
2025-07-17 01:19:07 -07:00
Daniel Han
ec8a4c6216
Update vision.py
2025-07-17 00:33:27 -07:00
Daniel Han
9d1afec41d
Update _utils.py
2025-07-14 02:45:19 -07:00
Daniel Han
85b1825e66
Merge branch 'main' into nightly
2025-07-14 02:44:42 -07:00
Roland Tannous
2d4ef50fb3
GRPO Fix - Support vllm pre-dequantized quantization states in fast_dequantize kernel ( #2943 )
...
* Support pre-dequantized quantization states in fast_dequantize kernel
* has_nested_quant conditional set to only
* Update utils.py
* Update utils.py
---------
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
2025-07-14 02:41:15 -07:00
Roland Tannous
497f8ccdc3
fix dataloader_num_workers value error in GRPOTrainer ( #2944 )
2025-07-14 01:43:33 -07:00
Muzammil Khan
20ef20422b
fix: change lora_dropout from int to float for type consistency ( #2949 )
...
Fixes "Argument of type 'float' cannot be assigned to parameter 'lora_dropout' of type 'int'" error by ensuring lora_dropout is consistently a float (0.0) rather than int (0) across vision.py, llama.py, and unsloth-cli.py
2025-07-14 01:42:07 -07:00
Datta Nimmaturi
382e9e03ce
Fix falcon H1 dropout issue ( #2938 )
...
Because we don't have down and gate multipliers, the MLP output values are too huge, causing NaN and unstable training. To bypass that lets rely on HF's implementation for the time being
2025-07-12 15:53:07 -07:00
DoubleMathew
ac8848b45f
patch falcon h1 inference ( #2932 )
2025-07-12 15:52:24 -07:00
Daniel Han
75e394b269
Update rl.py
2025-07-11 03:13:55 -07:00
Daniel Han
e69923e93f
Update rl.py
2025-07-11 03:11:07 -07:00
Daniel Han
a7c3f7e339
Merge branch 'main' into nightly
2025-07-11 03:04:28 -07:00
Daniel Han
751016d1f3
Uninitialized handler
2025-07-11 03:04:17 -07:00
Daniel Han
5ca7403870
Fixes
2025-07-11 00:01:37 -07:00
Daniel Han
ab4a69f5d8
Update llama.py
2025-07-10 17:18:40 -07:00
Daniel Han
82554bca7f
Merge branch 'main' into nightly
2025-07-10 17:12:59 -07:00
Daniel Han
9e5859581e
Fix GRPO
2025-07-10 17:12:49 -07:00
Michael Han
54916d9215
Merge pull request #2929 from rolandtannous/fix/fix-grpo-get-per-token-logps-argument-mismatch
...
Fix argument mismatch in GRPO _get_per_token_logps lambda function
2025-07-10 14:30:05 -07:00
Roland Tannous
4ea5249e65
Fix argument mismatch in GRPO _get_per_token_logps lambda function
2025-07-10 18:24:53 +00:00
Daniel Han
5c4fff6286
Merge branch 'main' into nightly
2025-07-10 07:03:59 -07:00
Daniel Han
8febcdf6e0
Many bug fixes ( #2927 )
...
* rename deepseek registration methods
* refactor naming for mistral and phi
* add global register models
* refactor model registration tests for new registry apis
* add model search method
* remove deprecated registration api
* add quant type test
* add registry readme
* make llama registration more specific
* clear registry when executing individual model registration file
* more registry readme updates
* Update _auto_install.py
* Llama4
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Synthetic data
* Update mapper.py
* Xet and Synthetic
* Update synthetic.py
* Update loader.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update pyproject.toml
* Delete .gitignore
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update _utils.py
* Update pyproject.toml
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update synthetic.py
* Update chat_templates.py
* Seasame force float16 / float32
* Fix Seasame
* Update loader.py
* Update vision.py
* Update vision.py
* Update vision.py
* Update loader.py
* is_multimodal
* Update loader.py
* Update loader.py
* Update loader.py
* Update loader.py
* Update vision.py
* Update vision.py
* Update vision.py
* UNSLOTH_DISABLE_STATIC_GENERATION
* Update vision.py
* Auto vision detection
* Sesame
* Whisper
* Update loader.py
* Update loader.py
* Update loader.py
* Update mapper.py
* Update vision.py
* Update vision.py
* Update vision.py
* Update vision.py
* Update vision.py
* Update vision.py
* Update loader.py
* Update loader.py
* Update loader.py
* Update loader.py
* Update _utils.py
* Update rl.py
* versioning
* Update rl.py
* Update rl.py
* Update rl.py
* Update rl.py
* Update rl.py
* logging
* Update pyproject.toml
* Update rl.py
* versioning
* Update rl.py
* Update rl.py
* Update rl_replacements.py
* Update rl_replacements.py
* Update rl.py
* Update rl_replacements.py
* Update rl_replacements.py
* logits / temperature
* Update rl_replacements.py
* Update pyproject.toml
* Update rl_replacements.py
* Update rl_replacements.py
* Debugging only
* Update llama.py
* Update llama.py
* Update rl_replacements.py
* Update rl_replacements.py
* Update rl_replacements.py
* Update rl_replacements.py
* Update rl_replacements.py
* Generic efficient GRPO
* Update rl_replacements.py
* Update rl_replacements.py
* Remove debugging
* Update rl_replacements.py
* Update rl_replacements.py
* Update vision.py
* Update llama.py
* Update rl_replacements.py
* versioning
* Update _utils.py
* Update vision.py
* Update mapper.py
* Update loader.py
* Update mapper.py
* Update vision.py
* Update loader.py
* Update vision.py
* Update loader.py
* Update _utils.py
* Update vision.py
* gradient checkpointing
* Gemma 3N fixes
* Update loader.py
* Versioning
* Gemma 3N fixes
* Update vision.py
* Update vision.py
* Update loader.py
* Update vision.py
* Fix setup.py
* setup.py
* Prints
* Update setup.py
* Update setup.py
* Update setup.py
* Update pyproject.toml
* Update pyproject.toml
* Update pyproject.toml
* Update pyproject.toml
* Update pyproject.toml
* Update pyproject.toml
* Update vision.py
* Update vision.py
* Update pyproject.toml
* Update vision.py
* Update _utils.py
* Update __init__.py
* Update __init__.py
* Small fixes
* Update vision.py
* Update vision.py
* versioning
* Update __init__.py
---------
Co-authored-by: jeromeku <jerome.ku@gmail.com>
Co-authored-by: Michael Han <107991372+shimmyshimmer@users.noreply.github.com>
2025-07-10 07:03:48 -07:00
Daniel Han
2efbe7b738
Update __init__.py
2025-07-10 07:03:28 -07:00
Daniel Han
1416c5015e
versioning
2025-07-10 07:01:44 -07:00
Daniel Han
7b73cfd916
Update vision.py
2025-07-10 05:15:03 -07:00
Daniel Han
ab51409dda
Update vision.py
2025-07-10 04:34:23 -07:00
Daniel Han
10cf55040a
Small fixes
2025-07-10 04:04:51 -07:00
Daniel Han
b7e96dde11
Merge branch 'main' into nightly
2025-07-10 04:02:01 -07:00
Datta Nimmaturi
a459558421
Move inputs to right devices. ( #2919 )
...
* Move tensors to right devices
* fix multi gpu for non mistral models
* multi GPU RoPE for gemma2
* Finish up multi GPU inference
* Make multiGPU rope a list
* Remove unnecessary transfer to CPU
* Remove unnecessary move to CPU
* Donot move inputs to device yet
will be handled separately in another PR
* Move inputs to appropriate decoder device
* Make device count global variable
* Cleanup RoPE device code
* Fixup num_gpu to device count
* Cleanup device counts
* Use device index for RoPE get_cache
* Donot typecast
* Use tuple instead of list for tensors. Use device index directly
* fixup move to device logic
2025-07-10 04:01:03 -07:00
Daniel Han
a766d8c43c
Merge branch 'main' into nightly
2025-07-10 01:50:14 -07:00
Daniel Han
267fa2ee70
Merge branch 'main' of https://github.com/unslothai/unsloth
2025-07-10 01:50:04 -07:00
Daniel Han
8017c098e6
Update llama.py
2025-07-10 01:50:03 -07:00
DoubleMathew
50f0481c08
if mlp doesn't exist in layer module check for feed_forward name for falcon h1 ( #2913 )
2025-07-09 23:29:41 -07:00
Daniel Han
9f2f2387ac
Update __init__.py
2025-07-09 16:30:57 -07:00
Datta Nimmaturi
5090fc3932
Explicitly check if xformers exists for attention ( #2889 )
2025-07-09 14:15:35 -07:00
Lei Zhenyuan
78a5e58932
[bugs] fix for casual mask ( #2868 )
...
* fix for casual mask
* use un_casual in sdpa
* add missing mask
* fix for type
2025-07-09 14:10:25 -07:00
Lei Zhenyuan
d188d5b7ce
add intel gpu with vllm support ( #2903 )
2025-07-09 14:08:38 -07:00
Datta Nimmaturi
dc9432b763
Dynamically adjust get_per_token_logps function and patch as well ( #2911 )
2025-07-09 14:07:33 -07:00
