Merge pull request #58 from huggingface/fix_stt_compile_mode

Assigning min new tokens to a compiled whisper graph on a thread brea…

STT/whisper_stt_handler.py

@@ -66,8 +66,9 @@ class WhisperSTTHandler(BaseHandler):
         if self.compile_mode not in (None, "default"):
             # generating more tokens than previously will trigger CUDA graphs capture
             # one should warmup with a number of generated tokens above max tokens targeted for subsequent generation
+            # hence, having min_new_tokens < max_new_tokens in the future doesn't make sense
             warmup_gen_kwargs = {
-                "min_new_tokens": self.gen_kwargs["min_new_tokens"],
+                "min_new_tokens": self.gen_kwargs["max_new_tokens"],  # Yes, assign max_new_tokens to min_new_tokens
                 "max_new_tokens": self.gen_kwargs["max_new_tokens"],
                 **self.gen_kwargs,
             }

arguments_classes/whisper_stt_arguments.py

@@ -33,12 +33,6 @@ class WhisperSTTHandlerArguments:
             "help": "The maximum number of new tokens to generate. Default is 128."
         },
     )
-    stt_gen_min_new_tokens: int = field(
-        default=0,
-        metadata={
-            "help": "The minimum number of new tokens to generate. Default is 0."
-        },
-    )
     stt_gen_num_beams: int = field(
         default=1,
         metadata={