fix temperature=0, flashinfer sample error

2026-05-27 08:26:02 +00:00 · 2025-04-07 12:30:47 +08:00
parent aac0c91d02
commit ec03bcbd7f
1 changed files with 2 additions and 0 deletions
--- a/ktransformers/server/backend/interfaces/balance_serve.py
+++ b/ktransformers/server/backend/interfaces/balance_serve.py
@@ -367,6 +367,8 @@ class BalanceServeInterface(BackendInterfaceBase):
        #@TODO add server
        stop_criteria =  [self.tokenizer.encode(self.tokenizer.eos_token, add_special_tokens=False),self.tokenizer.encode("<|im_end|>")]
        query_add.stop_criteria = stop_criteria
+        if temperature == 0:
+            temperature = 0.0001
        query_add.sample_options.temperature = temperature
        if top_p == 0:
            top_p = 0.0001