-
Notifications
You must be signed in to change notification settings - Fork 3k
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.
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
add flashmask rm #9154
add flashmask rm #9154
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## develop #9154 +/- ##
===========================================
- Coverage 53.06% 53.02% -0.05%
===========================================
Files 656 656
Lines 106147 106162 +15
===========================================
- Hits 56324 56288 -36
- Misses 49823 49874 +51 ☔ View full report in Codecov by Sentry. |
rejected_indexes = paddle.to_tensor( | ||
[[response_index[0], response_index[2]] for response_index in response_indexs] | ||
) | ||
chosen_hidden_states = hidden_states.gather_nd(chosen_indexes) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sequence parallel 等都不支持吗
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
适配了
"""main""" | ||
parser = PdArgumentParser((ModelArgument, DataArgument, TrainingArguments)) | ||
if len(sys.argv) == 2 and sys.argv[1].endswith(".json"): | ||
model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1])) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
这里要不支持下json和命令行,因为ce里面可能会增加命令行去覆盖json的配置,可参考这里:https://github.com/PaddlePaddle/PaddleNLP/blob/develop/llm/run_finetune.py#L77
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ok
logger.info("Start to create dataset") | ||
trans_func = partial(preprocess_preference_data, tokenizer=tokenizer, data_args=data_args, model_args=model_args) | ||
if data_args.lazy: | ||
zero_padding_dataset = ZeroPaddingIterableDataset |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
如果设置Lazy,那么按epoch保存会报错吗?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lazy只支持step
PR types
New features
PR changes
Others
Description
rm