update

llama_lora/globals.py

@@ -163,6 +163,7 @@ def load_gpu_info():
         preserve_loaded_models_count = math.floor(
             (available_cpu_ram * 0.8) / total_memory) - 1
         if preserve_loaded_models_count > 1:
+            ModelLRUCache = dynamic_import('.utils.model_lru_cache').ModelLRUCache
             print(
                 f"Will keep {preserve_loaded_models_count} offloaded models in CPU RAM.")
             Global.loaded_models = ModelLRUCache(preserve_loaded_models_count)

llama_lora/utils/eta_predictor.py

@@ -31,7 +31,7 @@ class ETAPredictor:
                 self._cleanup_old_data()
 
             # Only predict if we have enough data
-            if len(self.data) < 2 or self.data[-1][1] - self.data[0][1] < 5:
+            if len(self.data) < 2 or self.data[-1][1] - self.data[0][1] < 1:
                 return None
 
             first_step, first_time = self.data[0]