Simplify pytorch sampling kernel and logit processor (#2491)
This commit is contained in:
@@ -698,11 +698,6 @@ class ServerArgs:
|
||||
action="store_true",
|
||||
help="Disable Multi-head Latent Attention (MLA) for DeepSeek-V2.",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--disable-nan-detection",
|
||||
action="store_true",
|
||||
help="Disable the NaN detection for better performance.",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--disable-overlap-schedule",
|
||||
action="store_true",
|
||||
|
||||
Reference in New Issue
Block a user