tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

updating to fused (#2293)

Browse Source
This commit is contained in:
salman
2025-01-30 16:45:56 +00:00
committed by GitHub
parent 8779997ba5
commit ac471a697a
37 changed files with 51 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/e2e/patched/test_4d_multipack_llama.py
View File

@@ -52,7 +52,7 @@ class Test4dMultipackLlama(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,
@@ -96,7 +96,7 @@ class Test4dMultipackLlama(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,

2
tests/e2e/patched/test_fused_llama.py
View File

@@ -56,7 +56,7 @@ class TestFusedLlama(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 10,
"save_steps": 5,

4
tests/e2e/patched/test_llama_s2_attention.py
View File

@@ -56,7 +56,7 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 10,
"save_steps": 5,
@@ -96,7 +96,7 @@ class TestLlamaShiftedSparseAttention(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 10,
"save_steps": 5,

4
tests/e2e/patched/test_lora_llama_multipack.py
View File

@@ -61,7 +61,7 @@ class TestLoraLlama(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
}
)
@@ -116,7 +116,7 @@ class TestLoraLlama(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
}
)

4
tests/e2e/patched/test_mistral_samplepack.py
View File

@@ -55,7 +55,7 @@ class TestMistral(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,
@@ -96,7 +96,7 @@ class TestMistral(unittest.TestCase):
"gradient_accumulation_steps": 1,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch",
"optimizer": "adamw_torch_fused",
"lr_scheduler": "cosine",
"max_steps": 20,
"save_steps": 10,