[Model] Add support for MPT (#334)

This commit is contained in:
Woosuk Kwon
2023-07-03 16:47:53 -07:00
committed by GitHub
parent 7717d0838b
commit 404422f42e
11 changed files with 388 additions and 4 deletions
+1 -1
View File
@@ -12,7 +12,7 @@ from vllm import cache_ops
from vllm import pos_encoding_ops
from vllm.model_executor.input_metadata import InputMetadata
_SUPPORTED_HEAD_SIZES = [64, 80, 96, 128]
_SUPPORTED_HEAD_SIZES = [64, 80, 96, 112, 128]
class PagedAttention(nn.Module):