Switched back to google/gemma-3-4b-it

Gemma.py

@@ -16,12 +16,12 @@ class GemmaLLM:
         #     llm_int8_has_fp16_weight=False,
         # )
         
-        # model_id = "google/gemma-3-4b-it"
-        model_id = "google/gemma-3n-E4B-it-litert-preview"
+        model_id = "google/gemma-3-4b-it"
+        # model_id = "google/gemma-3n-E4B-it-litert-preview"
         # model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=quant_config)
         # tokenizer = AutoTokenizer.from_pretrained(model_id)
         
-        self.model = pipeline(model=model_id)
+        self.model = pipeline("text-generation", model=model_id)
         
     @spaces.GPU
     def generate(self, message) -> str: