recycle backend in simulstreaming thanks to new remove hooks function

whisperlivekit/simul_whisper/backend.py

@@ -7,6 +7,8 @@ from whisperlivekit.timed_objects import ASRToken, Transcript
 from whisperlivekit.warmup import load_file
 from whisperlivekit.simul_whisper.license_simulstreaming import SIMULSTREAMING_LICENSE
 from .whisper import load_model, tokenizer
+from .whisper.audio import TOKENS_PER_SECOND
+
 import os
 import gc
 logger = logging.getLogger(__name__)
@@ -21,6 +23,8 @@ except ImportError as e:
         """SimulStreaming dependencies are not available.
         Please install WhisperLiveKit using pip install "whisperlivekit[simulstreaming]".""")
 
+# TOO_MANY_REPETITIONS = 3
+
 class SimulStreamingOnlineProcessor:
     SAMPLING_RATE = 16000
 
@@ -33,33 +37,41 @@ class SimulStreamingOnlineProcessor:
         self.asr = asr
         self.logfile = logfile
         self.is_last = False
-        self.beg = 0.0
         self.end = 0.0
         self.cumulative_audio_duration = 0.0
         
         self.committed: List[ASRToken] = []
         self.last_result_tokens: List[ASRToken] = []
-        model = asr.get_new_model_instance()
-        self.model = PaddedAlignAttWhisper(
-            cfg=asr.cfg,
-            loaded_model=model)
+        self.load_new_backend()
         if asr.tokenizer:
             self.model.tokenizer = asr.tokenizer
 
-    def insert_audio_chunk(self, audio: np.ndarray, audio_stream_end_time: Optional[float] = None):
+    def load_new_backend(self):
+        model = self.asr.get_new_model_instance()
+        self.model = PaddedAlignAttWhisper(
+            cfg=self.asr.cfg,
+            loaded_model=model)
+
+    def insert_silence(self, silence_duration):
+        """
+        If silences are > 5s, we do a complete context clear. Otherwise, we just insert a small silence and shift the last_attend_frame
+        """
+        if silence_duration < 5:
+            gap_silence = torch.zeros(int(16000*min(silence_duration, 1.0)))
+            self.model.insert_audio(gap_silence)
+            self.model.last_attend_frame += int(TOKENS_PER_SECOND * (min(silence_duration, 1.0) - 1.0))
+        else:
+            self.model.refresh_segment(complete=True)
+            self.model.last_attend_frame += int(TOKENS_PER_SECOND * silence_duration)
+
+
+        
+    def insert_audio_chunk(self, audio: np.ndarray, audio_stream_end_time):
         """Append an audio chunk to be processed by SimulStreaming."""
             
         # Convert numpy array to torch tensor
         audio_tensor = torch.from_numpy(audio).float()
-        
-        # Update timing
-        chunk_duration = len(audio) / self.SAMPLING_RATE
-        self.cumulative_audio_duration += chunk_duration
-        
-        if audio_stream_end_time is not None:
-            self.end = audio_stream_end_time
-        else:
-            self.end = self.cumulative_audio_duration            
+        self.end = audio_stream_end_time #Only to be aligned with what happens in whisperstreaming backend.
         self.model.insert_audio(audio_tensor)
 
     def get_buffer(self):
@@ -81,7 +93,7 @@ class SimulStreamingOnlineProcessor:
         frames = [p["most_attended_frames"][0] for p in pr]
         tokens = tokens.copy()
         ret = []
-        for sw,st in zip(split_words,split_tokens):
+        for sw,st in zip(split_words, split_tokens):
             b = None
             for stt in st:
                 t,f = tokens.pop(0), frames.pop(0)
@@ -116,7 +128,29 @@ class SimulStreamingOnlineProcessor:
                     probability=0.95  # fake prob. Maybe we can extract it from the model?
                 )
                 new_tokens.append(token)
-                self.committed.extend(new_tokens)
+                
+            # identical_tokens = 0
+            # n_new_tokens = len(new_tokens)
+            # if n_new_tokens:
+            
+            self.committed.extend(new_tokens)
+            
+            # if token in self.committed:
+            #     pos = len(self.committed) - 1 - self.committed[::-1].index(token)
+            # if pos:
+            #     for i in range(len(self.committed) - n_new_tokens, -1, -n_new_tokens):
+            #         commited_segment = self.committed[i:i+n_new_tokens]
+            #         if commited_segment == new_tokens:
+            #             identical_segments +=1
+            #             if identical_tokens >= TOO_MANY_REPETITIONS:
+            #                 logger.warning('Too many repetition, model is stuck. Load a new one')
+            #                 self.committed = self.committed[:i]
+            #                 self.load_new_backend()
+            #                 return [], self.end
+
+            # pos = self.committed.rindex(token)
+
+            
             
             return new_tokens, self.end
 
@@ -137,10 +171,11 @@ class SimulStreamingOnlineProcessor:
 
     def __del__(self):
         # free the model and add a new model to stack.
-        del self.model
+        # del self.model
         gc.collect()
         torch.cuda.empty_cache()
-        self.asr.new_model_to_stack()
+        # self.asr.new_model_to_stack()
+        self.model.remove_hooks()
 
 class SimulStreamingASR():
     """SimulStreaming backend with AlignAtt policy."""
@@ -231,6 +266,7 @@ class SimulStreamingASR():
             self.models.append(self.load_model())
         new_model = self.models.pop()
         return new_model
+        # self.models[0]
 
     def new_model_to_stack(self):
         self.models.append(self.load_model())

whisperlivekit/simul_whisper/simul_whisper.py

@@ -56,6 +56,7 @@ class PaddedAlignAttWhisper:
         self.max_text_len = self.model.dims.n_text_ctx
         self.num_decoder_layers = len(self.model.decoder.blocks)
         self.cfg = cfg
+        self.l_hooks = []
 
         # model to detect end-of-word boundary at the end of the segment
         self.CIFLinear, self.always_fire, self.never_fire = load_cif(cfg,
@@ -69,7 +70,8 @@ class PaddedAlignAttWhisper:
             t = F.softmax(net_output[1], dim=-1)
             self.dec_attns.append(t.squeeze(0))
         for b in self.model.decoder.blocks:
-            b.cross_attn.register_forward_hook(layer_hook)
+            hook = b.cross_attn.register_forward_hook(layer_hook)
+            self.l_hooks.append(hook)
         
         self.kv_cache = {}
         def kv_hook(module: torch.nn.Linear, _, net_output: torch.Tensor):
@@ -82,10 +84,13 @@ class PaddedAlignAttWhisper:
             return self.kv_cache[module.cache_id] 
 
         for i,b in enumerate(self.model.decoder.blocks):
-            b.attn.key.register_forward_hook(kv_hook)
-            b.attn.value.register_forward_hook(kv_hook)
-            b.cross_attn.key.register_forward_hook(kv_hook)
-            b.cross_attn.value.register_forward_hook(kv_hook)
+            hooks = [
+                b.attn.key.register_forward_hook(kv_hook),
+                b.attn.value.register_forward_hook(kv_hook),
+                b.cross_attn.key.register_forward_hook(kv_hook),
+                b.cross_attn.value.register_forward_hook(kv_hook),
+            ]
+            self.l_hooks.extend(hooks)
 
         self.align_source = {}
         self.num_align_heads = 0
@@ -139,6 +144,11 @@ class PaddedAlignAttWhisper:
             self.inference.kv_cache = self.kv_cache
 
             self.token_decoder = BeamSearchDecoder(inference=self.inference, eot=self.tokenizer.eot, beam_size=cfg.beam_size)
+            
+    def remove_hooks(self):
+        print('remove hook')
+        for hook in self.l_hooks:
+            hook.remove()
 
     def create_tokenizer(self, language=None):
         self.tokenizer = tokenizer.get_tokenizer(