updates model catalog functions

llmware/models.py

@@ -312,6 +312,10 @@ def add_model_catalog_vars(cls, new_attr):
         cls.model_catalog_state_attributes.append(new_attr)
         return True
 
+    @classmethod
+    def reset_to_default_catalog(cls):
+        cls.registered_models = global_model_repo_catalog_list
+
 
 def pull_model_from_hf(model_card, local_model_repo_path, api_key=None, **kwargs):
 
@@ -503,6 +507,12 @@ def pull_latest_manifest(self):
         # will add to check manifest in global repo and make available for pull down
         return 0
 
+    def reset_to_default_catalog(self):
+        """ Resets model catalog to default list in model_configs """
+
+        _ModelRegistry().reset_to_default_catalog()
+        self.global_model_list = _ModelRegistry().get_model_list()
+
     def save_model_registry(self, fp=None, fn="llmware_model_catalog.json"):
 
         """ Utility method to export global model list to json file """
@@ -3933,14 +3943,6 @@ def __init__(self, model=None, tokenizer=None, model_name=None, api_key=None, mo
 
         self.get_token_counts = OVConfig().get_config("get_token_counts")
 
-        #   check for llmware path & create if not already set up
-        if not os.path.exists(LLMWareConfig.get_llmware_path()):
-            # if not explicitly set up by user, then create folder directory structure
-            LLMWareConfig.setup_llmware_workspace()
-
-        if not os.path.exists(LLMWareConfig.get_model_repo_path()):
-            os.mkdir(LLMWareConfig.get_model_repo_path())
-
         # please note that the external tokenizer is used solely for producing
         # input and output token counts - and can be switched off in OVConfig
         if self.get_token_counts:
@@ -10072,7 +10074,6 @@ def __init__(self, model_name=None, model_card=None, use_gpu_if_available=True,
 
         #   set verbose level in environ level - will be picked up by callback in whisper_cpp
         os.environ["whisper_cpp_verbose"] = GGUFConfigs().get_config("whisper_cpp_verbose")
-
         self.WHISPER_SR = GGUFConfigs().get_config("whisper_sr")
         self.strategy = GGUFConfigs().get_config("whisper_strategy")
         self.n_threads = GGUFConfigs().get_config("whisper_threads")
@@ -10154,7 +10155,6 @@ def load_model_for_inference(self, model_repo_path, model_card = None, **kwargs)
 
         # set to True by default - will display in 'real-time' the transcription
         self.params.print_realtime = GGUFConfigs().get_config("whisper_cpp_realtime_display")
-
         self.params.print_timestamps = True
         self.params.tdrz_enable = self.tiny_diarize
         self.params.progress_callback = whisper_progress_callback(self.callback)
@@ -10316,7 +10316,8 @@ def inference(self, prompt, inference_dict=None):
                 self.remove_segment_markers = inference_dict["remove_segment_markers"]
 
         #   preview before starting inference
-        self.preview()
+
+        # self.preview()
 
         #   note: updated dependencies for improved efficiency
         #   previously, used librosa library
@@ -10438,6 +10439,8 @@ def _generate(self, data):
 
         """ Executes lib_whisper generation on data from audio file. """
 
+        print("self.context len - ",self.context)
+
         w = self._lib.whisper_full(ctypes.c_void_p(self.context),
                                    self.params,
                                    data.ctypes.data_as(ctypes.POINTER(ctypes.c_float)),