Goekdeniz-Guelmez
e88f0fad4b
clean up
2025-03-09 00:18:33 +01:00
Goekdeniz-Guelmez
0bc2a881ad
generation should be fixed now
2025-03-09 00:16:40 +01:00
Gökdeniz Gülmez
46d6146102
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-03-08 22:41:10 +01:00
Awni Hannun
d2e02b3aae
fix mixed quant option ( #1326 )
2025-03-07 08:35:48 -08:00
Awni Hannun
595f5da146
remove lm head if unused ( #1324 )
2025-03-06 15:35:47 -08:00
cavit99
877d2a345b
Change DEFAULT_SEED to None for stochastic generation by default ( #1323 )
...
* Change DEFAULT_SEED to None for stochastic generation by default
* Update llms/mlx_lm/chat.py
* Update llms/mlx_lm/generate.py
---------
Co-authored-by: Awni Hannun <awni.hannun@gmail.com>
2025-03-06 06:49:35 -08:00
Awni Hannun
32d10036de
fix flaky test ( #1322 )
2025-03-05 14:00:09 -08:00
Gökdeniz Gülmez
e150621095
Adding multiple optimizers to mlx lm ( #1315 )
...
* initial commmit
* adding more customized YAML configuartion
* update YAML example file
* Changed the switch to set opt_class
* removing muon
* using default arguments
* udpate
2025-03-05 13:54:54 -08:00
Gökdeniz Gülmez
56d2db23e1
adding OLMoE architecture ( #1321 )
...
* initial commit
* udpate ACKNOWLEDGMENTS.md
* adding olmoe to training
* clean up
* faster generation
* remove sanitize method
* more clean ups
* adding SwitchGLU
* clean up
* a little faster and adding norm_topk_prob
* formated
2025-03-05 13:46:06 -08:00
Angelos Katharopoulos
e7267d30f8
Distributed support cifar ( #1301 )
2025-03-05 13:33:15 -08:00
Goekdeniz-Guelmez
f13a0d04ca
seperate functions
2025-03-05 15:28:12 +01:00
Goekdeniz-Guelmez
d723ddfeda
updates
2025-03-05 14:49:56 +01:00
Goekdeniz-Guelmez
9a36452519
updates
2025-03-05 14:42:34 +01:00
Goekdeniz-Guelmez
326935be49
updates
2025-03-05 14:40:23 +01:00
Goekdeniz-Guelmez
2d2f39f96e
updates
2025-03-05 14:25:55 +01:00
Goekdeniz-Guelmez
1f89453295
eos token return fix
2025-03-05 14:00:51 +01:00
Goekdeniz-Guelmez
2bde97fe13
minor speed improvement
2025-03-05 13:55:24 +01:00
Goekdeniz-Guelmez
3dfb21267b
updates
2025-03-05 12:59:41 +01:00
Awni Hannun
f621218ff5
Tool use example ( #1316 )
...
* tool use example
* nits
2025-03-04 13:53:20 -08:00
Awni Hannun
65aa2ec849
use a bool mask for attention ( #1319 )
2025-03-04 12:47:32 -08:00
Gökdeniz Gülmez
c817743333
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-03-03 22:13:42 +01:00
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
Goekdeniz-Guelmez
132225a018
updates
2025-03-01 22:23:33 +01:00
Goekdeniz-Guelmez
925e11439b
updates
2025-02-28 22:07:24 +01:00
Gökdeniz Gülmez
80e10a59d7
Merge branch 'main' into adding-GRPO-training
2025-02-28 21:16:02 +01:00
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
Goekdeniz-Guelmez
15d53279ae
batching fix
2025-02-28 16:02:40 +01:00
Gökdeniz Gülmez
a04eb02257
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-02-28 11:18:32 +01:00
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
Gökdeniz Gülmez
f27ed26b32
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-02-27 11:23:20 +01:00
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
Goekdeniz-Guelmez
fab2dc2688
smoll fix
2025-02-26 15:21:57 +01:00
Gökdeniz Gülmez
ef6ff92add
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-02-25 13:26:22 +01:00
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
Goekdeniz-Guelmez
53185c7f3d
last update, gn
2025-02-24 22:20:07 +01:00
Goekdeniz-Guelmez
e4eac9c97b
adding custom system message integration in dataset, more opimizations (generates now faster, while same RAM usage), fix for the identical generatrions, seperated the reward functions into a seperate file.
2025-02-24 20:49:22 +01:00
Gökdeniz Gülmez
bd5f081ca5
Merge branch 'ml-explore:main' into adding-GRPO-training
2025-02-22 19:11:26 +01:00
Goekdeniz-Guelmez
c0bd89aa15
add usage in LORA.md
2025-02-22 17:29:06 +01:00
Goekdeniz-Guelmez
9705ed908e
fix wrong generation in train
2025-02-22 17:21:08 +01:00
Usama Ahmed
09b641aaa7
Fix FutureWarning in torch.load by setting weights_only=True ( #1295 )
2025-02-22 06:08:54 -08:00
Goekdeniz-Guelmez
d9c4c6e60c
clean up and readding temperature argument
2025-02-22 02:34:56 +01:00
Goekdeniz-Guelmez
d653371e3d
nits
2025-02-22 02:12:02 +01:00
Goekdeniz-Guelmez
235348c211
generation speed improvement in training too from 3 t/s to 15 t/s
2025-02-22 02:03:01 +01:00
Goekdeniz-Guelmez
79de353530
nits
2025-02-22 01:05:58 +01:00
Goekdeniz-Guelmez
c51b0a2715
fix
2025-02-22 00:21:47 +01:00
