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

make e2e tests a bit faster by reducing test split size (#2522) [skip ci]

Browse Source 
* [ci] make e2e tests a bit faster by reducing test split size

* use 10% split of alpaca dataset to speed up dataset loading/tokenization

* reduce gas 4->2 for most e2e tests

* increase val set size for packing
This commit is contained in:
Wing Lian
2025-04-12 07:24:43 -07:00
committed by GitHub
parent 51267ded04
commit de8a625dd7
24 changed files with 54 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/e2e/multigpu/solo/test_flex.py
View File

@@ -55,7 +55,7 @@ class TestPackedFlex:
],
"num_epochs": 1,
"micro_batch_size": 2,
"gradient_accumulation_steps": 4,
"gradient_accumulation_steps": 2,
"output_dir": temp_dir,
"learning_rate": 0.00001,
"optimizer": "adamw_torch_fused",

19
tests/e2e/multigpu/test_llama.py
View File

@@ -58,12 +58,13 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
"max_steps": 2,
"micro_batch_size": 1,
"gradient_accumulation_steps": 4,
"gradient_accumulation_steps": 2,
# "gradient_checkpointing": True,
"output_dir": temp_dir,
"learning_rate": 0.00001,
@@ -201,7 +202,7 @@ class TestMultiGPULlama:
"num_epochs": 1,
"max_steps": 2,
"micro_batch_size": 2,
"gradient_accumulation_steps": 4,
"gradient_accumulation_steps": 2,
# "gradient_checkpointing": True,
"output_dir": temp_dir,
"warmup_steps": 0,
@@ -279,7 +280,7 @@ class TestMultiGPULlama:
"num_epochs": 1,
"max_steps": 2,
"micro_batch_size": 2,
"gradient_accumulation_steps": 4,
"gradient_accumulation_steps": 2,
# "gradient_checkpointing": True,
"output_dir": temp_dir,
"warmup_steps": 0,
@@ -335,6 +336,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -398,7 +400,7 @@ class TestMultiGPULlama:
"sample_packing": True,
"pad_to_sequence_len": True,
"sequence_len": 1024,
"val_set_size": 0.01,
"val_set_size": 0.05,
"special_tokens": {
"pad_token": "<|endoftext|>",
},
@@ -406,6 +408,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -484,6 +487,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -565,7 +569,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:25%]",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -660,7 +664,7 @@ class TestMultiGPULlama:
"sample_packing": True,
"pad_to_sequence_len": True,
"sequence_len": 1024,
"val_set_size": 0.01,
"val_set_size": 0.05,
"special_tokens": {
"pad_token": "<|endoftext|>",
},
@@ -668,6 +672,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -741,6 +746,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -814,6 +820,7 @@ class TestMultiGPULlama:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,

2
tests/e2e/multigpu/test_ray.py
View File

@@ -45,6 +45,7 @@ class TestMultiGPURay:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,
@@ -103,6 +104,7 @@ class TestMultiGPURay:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,

1
tests/e2e/multigpu/test_sp.py
View File

@@ -40,6 +40,7 @@ class TestSequenceParallelism:
{
"path": "tatsu-lab/alpaca",
"type": "alpaca",
"split": "train[:10%]",
},
],
"num_epochs": 1,