Awni Hannun
65aa2ec849
use a bool mask for attention ( #1319 )
2025-03-04 12:47:32 -08:00
Pierre-Louis
1bc3476a46
chore(lora): Add real-time log buffering fix for nohup execution ( #1311 )
...
* chore(lora): Add real-time log buffering fix for nohup execution
Disable Python stdout buffering to ensure logs appear in nohup.out in real-time instead of only after script completion.
* chore(lora): remove python 3.7+ check
* chore(lora): running pre-commit hook
---------
Co-authored-by: Pierre-Louis Létoquart <randlgint@proton.me>
2025-03-03 06:12:33 -08:00
Shunta Saito
269faa5fa4
Fix plamo2 model to use rms_norm ( #1308 )
...
* Fix plamo2 model to use rms_norm and enable sliding window attention
* Fix missing variable
* Remove sliding window attention impl. cause it should be done by using RotatingKVCache
* Remove unused imports
2025-03-03 06:12:02 -08:00
Awni Hannun
845cd8c01e
support kimi + more options in chat mode ( #1312 )
2025-02-28 11:33:18 -08:00
Awni Hannun
b2108a0de6
Allow mask prompt in config ( #1314 )
2025-02-28 11:33:04 -08:00
madroid
eb73549631
Generate: Support Prefill Response ( #1299 )
...
* Generate: Support Prefill Prompt
python -m mlx_lm.generate \
--model mlx-community/DeepSeek-R1-Distill-Qwen-1.5B-4bit \
--prompt "hello" \
--prefill-prompt "<think>\n"
* Generate: rename prefill-prompt to prefill-response
* nits
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2025-02-27 07:44:00 -08:00
Awni Hannun
00a7379070
Fixes for phi4 mini ( #1305 )
2025-02-26 16:21:54 -08:00
Awni Hannun
0f240a4c7e
Use max tokens from options in mlx_lm evaluate ( #1302 )
2025-02-26 15:46:16 -08:00
Awni Hannun
56e60ad5a6
fix manage for new transformers ( #1304 )
2025-02-26 15:44:57 -08:00
Pedro Cuenca
b7f742ef56
Mixed quant recipes ( #1300 )
...
* Mixed 3/6 and 2/6 recipes based on Alex Barron's
* format / nits
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2025-02-26 11:32:36 -08:00
Shunta Saito
c37e26a1a3
Add plamo-2-1b model ( #1283 )
...
* Add pfnet/plamo-2-1b
* Fix cache.py to support non-top level layers
* Use mlx's BaseModelArgs
* Fix model
* Use sanitize()
* Remove unnecessary changes
* Add plamo2.py
* Apply formatter
* Fix some part
* Allow a cache obj defined externally
* Fix channel first weights to channel last for right use of MLX's conv1d
* Remove unused code part
* Give all inputs when it's the first time call of model
* Fix import
* Include .jsonl files to download from Huggingface hub
* Fix reference to layers
* Remove unnecessary code and add a test for plamo2
* Do not pass mask to prepare_inputs_for_generation
* Fix to use repeat instead of tile
* Add state property to PlamoCache
* Add __iter__ and __next__ methods to PlamoCache
* cleanup
* cleanup
* fix
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2025-02-24 19:24:43 -08:00
Usama Ahmed
09b641aaa7
Fix FutureWarning in torch.load by setting weights_only=True ( #1295 )
2025-02-22 06:08:54 -08:00
Awni Hannun
3d793ecf68
Fix logits processor bugs with spec dec ( #1291 )
...
* Fix logits processor bugs with spec dec
* bump patch
2025-02-20 15:55:55 -08:00
Awni Hannun
85669451d0
Fix num layers in fine tune ( #1294 )
2025-02-20 13:32:01 -08:00
Awni Hannun
1cbf5cdac7
use more standard window strategy ( #1287 )
2025-02-19 06:22:51 -08:00
Matthias Neumayer
96bf37008e
Update README.md to include how to set temperature ( #1280 )
...
* Update README.md to include how to set temperature
* nits
---------
Co-authored-by: Awni Hannun <awni@apple.com>
2025-02-13 19:32:56 -08:00
Awni Hannun
7b07b14e67
add logits processor to spec gen ( #1260 )
2025-02-13 19:19:53 -08:00
Awni Hannun
ec30dc3538
hunyuan finetune ( #1270 )
2025-02-11 16:49:35 -08:00
Awni Hannun
42413c5d85
fix lora timings after validation ( #1278 )
2025-02-11 16:48:55 -08:00
Awni Hannun
f8cbf159e0
fix sharding for more even number of layers ( #1276 )
2025-02-11 16:26:59 -08:00
Awni Hannun
e879ea70e1
fix generation evaluations ( #1277 )
2025-02-11 16:10:30 -08:00
Matt Clayton
3d677f0870
Add "from_draft" to GenerationResponse ( #1272 )
...
* Add from_draft field in GenerationResponse
* Cleanup
* Re-work for minimal changes, add test
* Fix comment
2025-02-11 15:41:02 -08:00
Awni Hannun
bded1a8fcd
fix looping in whisper ( #1273 )
2025-02-10 13:04:35 -08:00
Chime Ogbuji
5865899c81
Completion only fine-tuning of instruction models with collections of HF datasets ( #1103 )
...
- Optional completion only fine-tuning with `--mask-prompt`
- Collections of Hugging Face datasets
---------
Co-authored-by: Awni Hannun <awni@apple.com>
2025-02-09 20:12:34 -08:00
Sri Harsha Pamu
1ced1b00ca
rm temp argument ( #1267 )
2025-02-09 11:39:11 -08:00
Awni Hannun
f58c7de901
Some improvements to speedup alignment computation in MLX Whisper ( #1259 )
...
* some improvements to speedup alignment computation in MLX Whisper
* fix alignment
2025-02-08 15:47:00 -08:00
Awni Hannun
1503bd4f55
support hunyuan 7b ( #1263 )
2025-02-08 15:46:47 -08:00
Awni Hannun
31611b62d7
Add IBM granite model ( #1265 )
...
* add granite
* add thinking option
2025-02-08 15:46:15 -08:00
Awni Hannun
6120a5f376
Faster DSv2/3 expert score computation ( #1257 )
...
* fix deepseek sharding (#1242 )
* compile and use put along axis in deep seek routing function
2025-02-07 10:24:57 -08:00
Awni Hannun
52c41b5b5a
Fix prompt cache for models without chat template ( #1250 )
...
* fix deepseek sharding (#1242 )
* fix prompt cache with no chat template
2025-02-06 11:10:58 -08:00
Nripesh Niketan
747c08e202
Chore: pre-commit bump ( #1253 )
2025-02-06 09:06:31 -08:00
Pedro Cuenca
e2e5478da5
READMEs: fix typo in link, minor update. ( #1246 )
2025-02-04 11:52:32 -08:00
Awni Hannun
21d0ab6e8a
fix deepseek sharding ( #1242 )
2025-02-03 16:59:50 -08:00
Gökdeniz Gülmez
0989c073b0
Optimizations for mamba1 ( #1213 )
...
* added mx.einsum() operations: before: 41.293 tokens-per-sec, after: 57.822 tokens-per-sec
* Fused Operations in delta, B, C = ... :. Before: 57.822 tokens-per-sec, after: 83.890 tokens-per-sec
* Pre-computing A_log. After: 83.890 tokens-per-sec, before: 85.848 tokens-per-sec
* Update MambaBlock, Batched Input Processing, Improved Cache Handling, Pre-computed Constants, Cleaner State Management, Explicit Return Values:. Before: 82.442 tokens-per-sec, after: 129.130 tokens-per-sec.
* cleaning up and adding apple copyright to helium modelfile
* update Copyright to this year
* nits + even faster
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2025-02-03 13:36:08 -08:00
Awni Hannun
d9924d08d1
Fix no validation in lora ( #1241 )
2025-02-03 09:55:24 -08:00
Awni Hannun
9c2ef38d4d
only download local shard ( #1240 )
2025-02-02 13:58:44 -08:00
Awni Hannun
e8afb59de4
better overflow correction ( #1229 )
2025-01-28 14:37:30 -08:00
Anchen
7a83077cd7
chore(mlx-lm): support text type content in messages ( #1225 )
...
* chore(mlx-lm): support text type content
* chore: optimize the messagef content processing
* nits + format
---------
Co-authored-by: Awni Hannun <awni@apple.com>
2025-01-27 17:13:50 -08:00
Awni Hannun
f44a52e2dc
batched min p and fix spec gen sampling ( #1222 )
2025-01-27 15:40:31 -08:00
Gökdeniz Gülmez
77faa14ba4
adding support for kyutai's helium ( #1208 )
...
* initial commit
* adding helium into training
* Update ACKNOWLEDGMENTS.md
* nits
* nits
* fixes / nits
---------
Co-authored-by: Awni Hannun <awni@apple.com>
2025-01-26 07:19:07 -08:00
Awni Hannun
9a3ddc3e65
some fixes for pipeline parallel deep seek r1 ( #1216 )
2025-01-21 19:40:29 -08:00
Victor Nogueira
df1406735b
Fix dataset variable name, in datasets.py ( #1212 )
2025-01-21 14:12:43 -08:00
Jarrett
07f88f8057
fix(lora): add back store_true default args ( #1205 )
2025-01-16 11:15:42 -08:00
Awni Hannun
50f0a7f6d9
add internlm3 ( #1206 )
2025-01-15 14:55:41 -08:00
Ivan Fioravanti
6ae6c72c2e
reduction moved to CPU in case of distributed training ( #1200 )
2025-01-14 17:20:42 -08:00
Awni Hannun
c117af83b8
fix gpt bigcode ( #1204 )
2025-01-13 10:22:32 -08:00
Chime Ogbuji
0228c46434
Custom local dataset features ( #1085 )
...
* Generalize prompt_feature and completion_feature for use in local datasets to facilitate compatibility with many other training dataset formats.
* Persist configured prompt/completion key
* rebase + nits
---------
Co-authored-by: Awni Hannun <awni@apple.com>
2025-01-13 10:01:18 -08:00
Prince Canuma
bf2da36fc6
Fix Cohere2: mask shape error (long context) ( #1202 )
...
* fix mask shape error (long context)
* Update llms/mlx_lm/models/cohere2.py
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
* revert layer_idx
* black formatting
* Update cohere2.py
* format
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
Co-authored-by: Awni Hannun <awni@apple.com>
2025-01-12 12:58:08 -08:00
Xingjun.Wang
514502da22
Support snapshot_download for ModelScope ( #1194 )
...
* add MLX_USE_MODELSCOPE env
* update
* update snapshot_download
* update
* remove modelscope dependency and add import check
* update
* nits
* fix
---------
Co-authored-by: wangxingjun778 <jason@U-C7X6TX5G-2239.local>
Co-authored-by: Awni Hannun <awni@apple.com>
2025-01-10 15:29:34 -08:00
Awni Hannun
93c5cfd781
Add a speculative decoding generator ( #1155 )
...
* add a speculative decoding generator
* fix
* fixes
* optional kwarg pop
2025-01-10 15:27:08 -08:00
