OLMo icon indicating copy to clipboard operation
OLMo copied to clipboard

Why does config/llama7.yaml not use OlmoLlamaBlock?

Open gahdritz opened this issue 1 year ago • 1 comments

A sample "LLaMA-like" config is provided, but it doesn't use the OlmoLlamaBlock BlockType defined in model/olmo.py. Why is that?

https://github.com/allenai/OLMo/blob/97296e610c24dd1bb098ec64660dfcafcba62d24/configs/llama7.yaml#L21

gahdritz avatar Feb 10 '24 06:02 gahdritz

The remaining changes to make that config more Llama-like in behavior are in https://github.com/allenai/OLMo/pull/358. I don't remember why we didn't merge that change, but the throughput was very poor due to how our implementation tried to get identical results in the attention mechanism.

2015aroras avatar Feb 13 '24 23:02 2015aroras

I apologize for our delay in response. In order to help surface current, unresolved issues, we are closing tickets prior to February 29. Please reopen your ticket if you are continuing to experience this issue. Thank you!

dumitrac avatar Apr 30 '24 18:04 dumitrac