Commit 7f59769

committed

Fix MoE crash and hang issues

Signed-off-by: Jevin Jiang <jevin0change@gmail.com>

1 parent 0b4594c commit 7f59769Copy full SHA for 7f59769

2 files changed

+259

-176

lines changed

tpu_inference
- kernels/fused_moe/v1
  - kernel.py
- layers/vllm/quantization
  - unquantized.py

+259

-176

lines changed

Comments

(0)