search

google / jetstream-pytorch

PyTorch/XLA integration with JetStream (https://github.com/google/JetStream) for LLM inference"
Apache License 2.0
32 stars 14 forks source link

Fix sharding config file name bug #86

Closed FanhaiLu1 closed 3 months ago

FanhaiLu1 commented 3 months ago

Fix sharding config file name bug. Two fixed:

  1. Current code uses llama2 or llama3 as file name, but the sharding yaml file name is llama.
  2. Engine has logic to handle none sharding_config, removed the duplicated logic in run server