excluded ollama imports when running in hf

repository/intel_npu.py

@@ -43,6 +43,7 @@ class IntelNpuRepository(Repository):
         input_ids = (self.tokenizer.apply_chat_template(self.get_message_history(), add_generation_prompt=True,
                                                         return_tensors="pt")
                      .to(self.model.device))
+
         outputs = self.model.generate(input_ids, eos_token_id=self.terminators, do_sample=True, max_new_tokens=2000, cache_position=None)
         generated_token_array = outputs[0][len(input_ids[0]):]
         generated_tokens = "".join(self.tokenizer.batch_decode(generated_token_array, skip_special_tokens=True))

repository/repository.py

@@ -3,7 +3,7 @@ from utils.env_utils import in_hf
 
 if not in_hf():
     from repository.intel_npu import IntelNpuRepository
-from repository.ollama import OllamaRepository
+    from repository.ollama import OllamaRepository
 from repository.ondemand import OndemandRepository
 from repository.repository_abc import Model
 from repository.testing_repo import TestingRepository
@@ -13,11 +13,11 @@ def get_repository(implementation: str, model: Model, system_msg: str = None, lo
     known_implementations = ["ollama", "intel_npu", "testing", "ondemand"]
     if not implementation or implementation.lower() not in known_implementations:
         raise ValueError(f"Unknown implementation {implementation}. Known implementations: {known_implementations}")
-    if "ollama" == implementation:
-        return OllamaRepository(model, system_msg)
     if not in_hf():
         if "intel_npu" == implementation:
             return IntelNpuRepository(model, system_msg, log_to_file)
+        if "ollama" == implementation:
+            return OllamaRepository(model, system_msg)
     if "ondemand" == implementation:
         return OndemandRepository(model, system_msg, log_to_file)
     if "testing" == implementation: