Xingtong Liu

icon indicating a website link https://www.linkedin.com/in/xingtong-liu-ph-d-b43b27131/ icon indicating an email [email protected]

icon company Intuitive Surgical, Inc. icon location Cupertino, California icon location Senior Machine Learning Engineer at Intuitive Surgical, Inc. | Ph.D. in Computer Science from Johns Hopkins University

Results 3 issues of Xingtong Liu

Plans on integrating qlora 4bit finetuning?

My understanding is currently the repo provides 4bit only for inference but not finetuning. If this is the case, is there a plan for integrating QLoRA-style 4bit finetuning?

falcon.cpp: tensor 'lm_head.weight' is missing from model

2 comment

Whenever I try to load a QLoRA-merged Falcon 40B model, the error below happens. `error loading model: falcon.cpp: tensor 'lm_head.weight' is missing from model` The hacky way I did to...

Seemingly missing reshaping operation in point prompt encoding in PromptEncoder

https://github.com/facebookresearch/segment-anything/blob/main/segment_anything/modeling/prompt_encoder.py#L81-L85 If I read it correctly, the shape of points input is `[B, N, 2]`, where B is the batch size and N is the number of points per image....