fixed tensor types

app.py

@@ -28,6 +28,7 @@ def load_model():
         model = AutoModelForCausalLM.from_pretrained(
             MODEL_NAME,
             trust_remote_code=True,
+            torch_dtype=torch.float16,
             device_map="auto",
             attn_implementation="eager",
             use_cache=True,

requirements.txt

@@ -1,5 +1,6 @@
 torch>=2.0.0,<2.3.0
 transformers==4.48.0
 accelerate>=0.20.0
+bitsandbytes>=0.41.0
 gradio>=4.0.0,<5.0.0
 safetensors>=0.4.0