Update app.py

app.py

@@ -48,11 +48,12 @@ def transliterate_to_sinhala(text):
 # conv_model_name = "microsoft/Phi-3-mini-4k-instruct"  # Use GPT-2 instead of the gated model
 # tokenizer = AutoTokenizer.from_pretrained(conv_model_name, trust_remote_code=True)
 # model = AutoModelForCausalLM.from_pretrained(conv_model_name, trust_remote_code=True).to(device)
+pipe1 = pipeline("text-generation", model="TinyLlama/TinyLlama-1.1B-Chat-v1.0", torch_dtype=torch.bfloat16, device_map="auto")
 
-client = InferenceClient("google/gemma-2b-it")
+# client = InferenceClient("google/gemma-2b-it")
 
 def conversation_predict(text):
-    return client.text_generation(text, return_full_text=False)
+    # return client.text_generation(text, return_full_text=False)
     # pipe = pipeline(
     # "text-generation",
     # model=model,
@@ -71,6 +72,9 @@ def conversation_predict(text):
     # outputs = model.generate(**input_ids)
     # return tokenizer.decode(outputs[0])
 
+    outputs = pipe1(text, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
+    return outputs[0]["generated_text"]
+
 def ai_predicted(user_input):
     if user_input.lower() == 'exit':
         return "Goodbye!"