Training in progress - step 364

asr_config.py

@@ -63,6 +63,10 @@ class ASRConfig(transformers.PretrainedConfig):
         lora_dropout: float = 0.0,
         lora_target_modules: Optional[list] = None,  # Default: all linear layers
         freeze_projector: bool = False,  # True for Stage 2 (LoRA-only training)
+        do_sample: bool = False,
+        temperature: Optional[float] = None,
+        top_p: Optional[float] = None,
+        top_k: Optional[int] = None,
         max_new_tokens: Optional[int] = None,
         min_new_tokens: Optional[int] = None,
         repetition_penalty: Optional[float] = None,
@@ -169,6 +173,10 @@ class ASRConfig(transformers.PretrainedConfig):
             else generation_defaults["no_repeat_ngram_size"]
         )
         self.use_cache = use_cache if use_cache is not None else generation_defaults["use_cache"]
+        self.do_sample = do_sample
+        self.temperature = temperature
+        self.top_p = top_p
+        self.top_k = top_k
 
         if "audio_config" not in kwargs:
             self.audio_config = transformers.AutoConfig.from_pretrained(audio_model_id)

asr_modeling.py

@@ -136,11 +136,11 @@ class ASRModel(PreTrainedModel, GenerationMixin):
         self.generation_config.max_new_tokens = config.max_new_tokens
         self.generation_config.min_new_tokens = config.min_new_tokens
         self.generation_config.num_beams = config.num_beams
-        self.generation_config.do_sample = False
-        # Clear sampling params (inherited from LLM) since we use greedy decoding
-        self.generation_config.temperature = None
-        self.generation_config.top_p = None
-        self.generation_config.top_k = None
+        self.generation_config.do_sample = config.do_sample
+        # Set sampling params from config (None means use model defaults)
+        self.generation_config.temperature = config.temperature
+        self.generation_config.top_p = config.top_p
+        self.generation_config.top_k = config.top_k
         self.generation_config.use_cache = config.use_cache
         self.generation_config.length_penalty = config.length_penalty
         self.generation_config.repetition_penalty = config.repetition_penalty
@@ -730,7 +730,7 @@ class ASRModel(PreTrainedModel, GenerationMixin):
             tokenize=True,
             add_generation_prompt=True,
             return_tensors="pt",
-            enable_thinking=False,
+            enable_thinking=False,  # Disable Qwen3 thinking mode for ASR
         ).to(device)
 
         if input_ids.dim() == 1:

diarization.py

@@ -737,7 +737,7 @@ class SpeakerDiarizer:
 
             cls._pyannote_pipeline = Pipeline.from_pretrained(
                 "pyannote/speaker-diarization-3.1",
-                use_auth_token=hf_token,
+                token=hf_token,
             )
             cls._pyannote_pipeline.to(torch.device(_get_device()))