final

Author: yousef-rafat

comfy/autoregressive_sampling.py

@@ -255,6 +255,8 @@ def __init__(self, model, device, kv_cache_lengths: list = [1024, 4096, 8192]):
         self.dtype = model.dtype
 
         self.model = model
+        self.model.generation_config = GenerationConfig.from_model_config(self.model.config)
+        self.model.generation_config.cache_implementation = self.model.cache_implementation
 
         text_config = self.model.cache_config
         self.cache_config = CacheConfig(
@@ -331,8 +333,6 @@ def generate(self, input_ids: Optional[torch.LongTensor] = None, max_new_length:
                                                 do_sample = do_sample,
                                                 temperature = temperature)
 
-        generation_config, kwargs = self._prepare_generation_config(generation_config, **kwargs)
-
         generation_config, model_kwargs = self._prepare_generation_config(
             generation_config, **kwargs
         )

comfy/ldm/higgsv2/model.py

@@ -428,14 +428,13 @@ def __init__(self, device = None, dtype = None, operations = None, **kwargs):
         self.cache_config = kwargs["text_config"]
         self.hidden_dim = kwargs["text_config"]["hidden_size"]
         self.max_seq_len = kwargs["text_config"]["max_position_embeddings"]
+        self.cache_implementation = "static"
         self.use_kv_buckets = kwargs.get("use_kv_buckets", False)
 
         self.dtype = dtype
         self.device = device
         self.config = kwargs
 
-        self.generation_config = GenerationConfig.from_model_config(kwargs)
-        self.generation_config.cache_implementation = self.cache_implementation = "static"
 
         self.audio_out_bos_token_id = 128013
         self.audio_eos_token_id = 128012

comfy_extras/nodes_audio.py

@@ -174,8 +174,6 @@ def convert_to_ml_format(self, clip, text, audio=None):
         current_role = None
         collecting_system = False
         system_buffer = []
-        collecting_instruction = False
-        instruction_buffer = []
 
         for line in lines:
             line = line.strip()
@@ -198,26 +196,6 @@ def convert_to_ml_format(self, clip, text, audio=None):
                     collecting_system = False
                 continue
 
-            # generation instruction start
-            if "<|generation_instruction_start|>" in line:
-                collecting_instruction = True
-                instruction_buffer = []
-                continue
-
-            if collecting_instruction:
-                if "<|generation_instruction_end|>" in line:
-                    instruction_text = "\n".join(instruction_buffer)
-                    # include both start and end tokens
-                    messages.append(Message(
-                        role="user",
-                        content=f"<|generation_instruction_start|>\n{instruction_text}\n<|generation_instruction_end|>"
-                    ))
-                    instruction_buffer = []
-                    collecting_instruction = False
-                else:
-                    instruction_buffer.append(line)
-                continue
-
             # speaker lines SPEAKER-0: text
             match = re.match(r"SPEAKER-(\d+):\s*(.*)", line, re.IGNORECASE)
             if match: