luckyyangrun
luckyyangrun
Are there any plans to fix it?
I tried the code in this CR, and unfortunately, I found that the loss does not decrease during training.
1.5B model even with same model
> Hi! This step constructs these files. Thanks for your reply! After reading through script_build.py, I believe this script is responsible for reading the following preprocessed files: kv_store_text_chunks.json kv_store_entities.json kv_store_hyperedges.json...