Skip to content

remove flashmask checker #2631

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Sep 18, 2025
Merged

remove flashmask checker #2631

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
24a2c3e
remove flashmask checker
WYB27 Sep 18, 2025
5cdf24f
remove flashmask checker
WYB27 Sep 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 0 additions & 42 deletions examples/run_finetune.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,26 +37,8 @@
AutoModelForCausalLM,
AutoModelForCausalLMPipe,
AutoTokenizer,
DeepseekV2ForCausalLM,
DeepseekV2ForCausalLMPipe,
DeepseekV3ForCausalLM,
DeepseekV3ForCausalLMPipe,
Ernie4_5_MoeForCausalLM,
Ernie4_5_MoeForCausalLMPipe,
Ernie4_5ForCausalLM,
Ernie4_5ForCausalLMPipe,
Llama3Tokenizer,
LlamaForCausalLM,
LlamaForCausalLMPipe,
LlamaTokenizer,
Qwen2ForCausalLM,
Qwen2ForCausalLMPipe,
Qwen2MoeForCausalLM,
Qwen2MoeForCausalLMPipe,
Qwen3ForCausalLM,
Qwen3ForCausalLMPipe,
Qwen3MoeForCausalLM,
Qwen3MoeForCausalLMPipe,
)
from paddleformers.transformers.configuration_utils import LlmMetaConfig
from paddleformers.trl import DataConfig, ModelConfig, SFTConfig, SFTTrainer
Expand All @@ -66,27 +48,6 @@
# Fine-tune Environment Variables to support sharding stage1 overlap optimization.
os.environ["USE_CASUAL_MASK"] = "False"

flash_mask_support_list = [
DeepseekV2ForCausalLM,
DeepseekV2ForCausalLMPipe,
DeepseekV3ForCausalLM,
DeepseekV3ForCausalLMPipe,
Ernie4_5ForCausalLM,
Ernie4_5ForCausalLMPipe,
Ernie4_5_MoeForCausalLM,
Ernie4_5_MoeForCausalLMPipe,
LlamaForCausalLM,
LlamaForCausalLMPipe,
Qwen2ForCausalLM,
Qwen2ForCausalLMPipe,
Qwen2MoeForCausalLM,
Qwen2MoeForCausalLMPipe,
Qwen3ForCausalLM,
Qwen3ForCausalLMPipe,
Qwen3MoeForCausalLM,
Qwen3MoeForCausalLMPipe,
]


def main():
parser = PdArgumentParser((ModelConfig, DataConfig, SFTConfig))
Expand Down Expand Up @@ -192,9 +153,6 @@ def main():
else:
model = model_class.from_config(model_config, dtype=dtype)

Copy link
Collaborator

@lugimzzz lugimzzz Sep 18, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

去掉 flash_mask_support_list内容和相关import

if model_args.attn_impl == "flashmask" and not any(isinstance(model, cls) for cls in flash_mask_support_list):
raise NotImplementedError(f"{model.__class__} not support flash mask.")

if training_args.do_train and model_args.neftune:
# Inspired by https://github.com/neelsjain/NEFTune
if hasattr(model, "get_input_embeddings"):
Expand Down