jiaqiw09

add NPU support for huggingface.py

2

## what this PR do issue: #1797 This PR add NPU support for huggingface.py. It just does some fix of existing code to support NPU device. ## what part to...

Add NPU support for huggingface.py

1

## Feature request It would be nice if the lm_eval could be used directly on the Ascend NPU without modifying the source code of `huggingface.py`. ## Motivation In China, Ascend...

# What does this PR do? Aggregated RuntimeArgs Container - Introduced RuntimeArgs dataclass that aggregates DataArguments, ModelArguments, TrainingArguments, and SampleArguments - Supports partial aggregation (e.g., only model+training for training scenarios)...

[AutoTP + DS2] no memory reduction when using auto = 4

4

I trained model qwen2.5-7B in GPU with llamafactory deepspeed zero2 + autotp. And there is no obvious memory reduction. When individually using zero2, the average memory of 8 cards is...