syspider

我打断点看了下，没有用到gpt2的past_key_values机制，博主后面有空可以改进一下，应该能加速不少

How many A100s used for training?

1

Thank you for the great job! I am just a bit curious about the computing resources used for LLaVA-NeXT.

Question about M4-Instruct datasets

3

Thank your for your kindly release! But when i looking at the annotations of M4-Instruct, the FIRST sample just quite confused me. Here is the snapshot:  The human and...

Question about target_ids in finetune.py

4

 Normally system message and user's instruction are not calculated into loss, which can be done by some process like replacing the corresponding ids to ignore_token_id in target_ids. But in...