github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-02 01:28:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

[ci] fix shardformer tests. (#5255)

Browse Source 
* fix ci

fix

* revert: revert p2p

* feat: add enable_metadata_cache option

* revert: enable t5 tests

---------

Co-authored-by: Wenhao Chen <cwher@outlook.com>
This commit is contained in:
flybird11111
2024-01-11 19:07:45 +08:00
committed by GitHub
parent 756c400ad2
commit e830ef917d
4 changed files with 20 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/test_shardformer/test_model/test_shard_gpt2.py
View File

@@ -165,7 +165,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
)
@clear_cache_before_run()
def run_gpt2_test(test_config):
sub_model_zoo = model_zoo.get_sub_registry("transformers_gpt")
sub_model_zoo = model_zoo.get_sub_registry("transformers_gpt", exclude="transformers_gptj")
for name, (model_fn, data_gen_fn, output_transform_fn, loss_fn, _) in sub_model_zoo.items():
check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn, test_config)
@@ -200,7 +200,7 @@ def run_gpt2_test(test_config):
)
@clear_cache_before_run()
def run_gpt2_3d_test(test_config):
sub_model_zoo = model_zoo.get_sub_registry("transformers_gpt")
sub_model_zoo = model_zoo.get_sub_registry("transformers_gpt", exclude="transformers_gptj")
for name, (model_fn, data_gen_fn, output_transform_fn, loss_fn, _) in sub_model_zoo.items():
check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn, test_config)

6
tests/test_shardformer/test_model/test_shard_t5.py
View File

@@ -86,6 +86,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 2,
"enable_metadata_cache": False,
"enable_all_optimization": True,
"use_lazy_init": True,
"precision": "fp16",
@@ -95,6 +96,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 1,
"pp_size": 2,
"num_microbatches": 4,
"enable_metadata_cache": False,
"use_lazy_init": False,
"precision": "fp16",
"initial_scale": 1,
@@ -110,6 +112,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 1,
"pp_size": 4,
"num_microbatches": 4,
"enable_metadata_cache": False,
"enable_all_optimization": False,
"use_lazy_init": False,
"precision": "fp32",
@@ -128,6 +131,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 1,
"pp_size": 2,
"num_microbatches": 2,
"enable_metadata_cache": False,
"enable_all_optimization": True,
"use_lazy_init": True,
"zero_stage": 1,
@@ -159,6 +163,7 @@ def run_t5_test(test_config):
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 4,
"enable_metadata_cache": False,
"enable_all_optimization": False,
"use_lazy_init": False,
"precision": "fp32",
@@ -168,6 +173,7 @@ def run_t5_test(test_config):
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 4,
"enable_metadata_cache": False,
"enable_all_optimization": False,
"use_lazy_init": False,
"precision": "fp16",

5
tests/test_shardformer/test_model/test_shard_whisper.py
View File

@@ -114,6 +114,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 2,
"enable_metadata_cache": False,
"enable_all_optimization": True,
"use_lazy_init": True,
"precision": "fp32",
@@ -123,6 +124,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 1,
"pp_size": 2,
"num_microbatches": 4,
"enable_metadata_cache": False,
"use_lazy_init": False,
"precision": "fp32",
"initial_scale": 1,
@@ -138,6 +140,7 @@ def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn,
"tp_size": 1,
"pp_size": 4,
"num_microbatches": 4,
"enable_metadata_cache": False,
"use_lazy_init": False,
"precision": "fp32",
},
@@ -163,6 +166,7 @@ def run_whisper_test(test_config):
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 4,
"enable_metadata_cache": False,
"enable_all_optimization": False,
"use_lazy_init": False,
"precision": "fp32",
@@ -172,6 +176,7 @@ def run_whisper_test(test_config):
"tp_size": 2,
"pp_size": 2,
"num_microbatches": 2,
"enable_metadata_cache": False,
"enable_all_optimization": False,
"use_lazy_init": False,
"precision": "fp32",