search

Alpha-VLLM / LLaMA2-Accessory

An Open-source Toolkit for LLM Development
https://llama2-accessory.readthedocs.io/
Other
2.68k stars 170 forks source link

InternLM inference and training are problematic #78

Closed June01 closed 11 months ago

June01 commented 11 months ago

The following is my code to run the demo, with pretrained model, llama_config, tokenizer. However, the output of the network is chaos even with a simple Question "hello". Could you please look into it?

torchrun --nproc-per-node=1  demos/single_turn.py \
--llama_config /path/to/params.json --tokenizer_path /path/to/tokenizer.model \
--pretrained_path /path/to/alpaca_finetuned

image

ChrisLiu6 commented 11 months ago

You are now loading internlm checkpoint to a llama model, Please try adding --llama_type internlm.

June01 commented 11 months ago

Thanks! It works.

On Tue, 26 Sept 2023 at 18:41, Chris Liu @.***> wrote:

You are now loading internlm checkpoint to a llama model, Please try adding --llama_type internlm.

— Reply to this email directly, view it on GitHub https://github.com/Alpha-VLLM/LLaMA2-Accessory/issues/78#issuecomment-1735281389, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC5TQGHNEZVCK3NT4ADTAP3X4KWN5ANCNFSM6AAAAAA5HIE6Y4 . You are receiving this because you authored the thread.Message ID: @.***>