Mistral Experimental org
edited 18 days ago
Optimized .pte files

- Model size is 4.4GB (before 6GB). bf16 accumulation
- Fixed a bug in preprocessor.pte which was allocating 6GB of RAM, wastefully.

- Also the latest main contains optimizations such as init time improvements.
Mistral Experimental org

Thanks!

patrickvonplaten changed pull request status to merged

Sign up or log in to comment