Add Post_model_load, post_lora_load, post_train, post_train_unload function calls (#2539)

* Update train.py

add post_model_load and post_lora_load model calss.

* Update train.py

add post_train and post_train_unload function calls

* Update train.py

* Update base.py

* Update train.py

* chore: lint

* clarify plugin hooks

* Update src/axolotl/integrations/base.py

Co-authored-by: Dan Saunders <danjsaund@gmail.com>

* Update src/axolotl/utils/models.py

Co-authored-by: Dan Saunders <danjsaund@gmail.com>

* Update src/axolotl/utils/models.py

Co-authored-by: Dan Saunders <danjsaund@gmail.com>

* Update src/axolotl/integrations/base.py

Co-authored-by: Dan Saunders <danjsaund@gmail.com>

* Update models.py

* Update models.py

* remove extra call to post_model_load

* chore: lint

* add test for hooks and gc trainer

* disable duplicated code check for test

* fix the path and add better handling

---------

Co-authored-by: Wing Lian <wing@axolotl.ai>
Co-authored-by: Dan Saunders <danjsaund@gmail.com>
This commit is contained in:
divyanshuaggarwal
2025-04-28 19:40:28 +05:30
committed by GitHub
parent 5d182a1056
commit 170cdb5be9
6 changed files with 245 additions and 10 deletions

View File

@@ -48,13 +48,13 @@ class TestLoraLlama(unittest.TestCase):
},
],
"num_epochs": 1,
"micro_batch_size": 8,
"micro_batch_size": 2,
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 20,
"max_steps": 5,
}
)