Misc fix for min_p_sampling, --cuda-graph-bs (#2761)

This commit is contained in:
Lianmin Zheng
2025-01-07 02:52:53 -08:00
committed by GitHub
parent 6d08ce2aa9
commit bdc1acf6cd
17 changed files with 135 additions and 63 deletions

View File

@@ -117,6 +117,11 @@ class LogitsProcessor(nn.Module):
self.final_logit_softcapping = getattr(
self.config, "final_logit_softcapping", None
)
if (
self.final_logit_softcapping is not None
and self.final_logit_softcapping < 0
):
self.final_logit_softcapping = None
def forward(
self,