Update app.py

app.py

@@ -45,12 +45,13 @@ h1 {
 """
 
 # Load the tokenizer and model
-tokenizer = AutoTokenizer.from_pretrained("CyberNative-AI/Colibri_8b_v0.1_gptq_128_4bit")
-model = AutoGPTQForCausalLM.from_quantized("CyberNative-AI/Colibri_8b_v0.1_gptq_128_4bit", device="cpu")
+with torch.autocast(device_type='cpu'):
+    tokenizer = AutoTokenizer.from_pretrained("CyberNative-AI/Colibri_8b_v0.1_gptq_128_4bit")
+    model = AutoGPTQForCausalLM.from_quantized("CyberNative-AI/Colibri_8b_v0.1_gptq_128_4bit", dtype=torch.bfloat16, device="cpu")
 
 terminators = [
     tokenizer.eos_token_id,
-    tokenizer.convert_tokens_to_ids("<|eot_id|>")
+    tokenizer.convert_tokens_to_ids("<|im_end|>")
 ]
 
 @spaces.GPU(duration=120)