f

app.py

@@ -28,7 +28,6 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 # For a standard FP16 or FP32 load (no bitsandbytes):
 model = AutoModelForCausalLM.from_pretrained(
     "cheberle/autotrain-35swc-b4r9z",
-    device_map="auto",           # Device automatically mapped across GPUs or CPU
     torch_dtype=torch.float16,    # Or "auto", or float32
     trust_remote_code=True
 )