Upload folder using huggingface_hub

api.py

@@ -3,7 +3,6 @@ from flask import Flask, request, jsonify
 from PIL import Image
 from io import BytesIO
 import base64
-import torch
 from transformers import AutoProcessor, AutoModelForCausalLM
 import threading
 
@@ -16,20 +15,9 @@ parser.add_argument('--device', type=str, choices=['cpu', 'gpu'], default='auto'
 args = parser.parse_args()
 
 # Determine the device
-if args.device == 'auto':
-    device = "cuda:0" if torch.cuda.is_available() else "cpu"
-elif args.device == 'gpu':
-    if torch.cuda.is_available():
-        device = "cuda:0"
-    else:
-        raise ValueError("GPU option specified but no GPU is available.")
-else:
-    device = "cpu"
-
-torch_dtype = torch.float16 if device.startswith("cuda") else torch.float32
-
+device = "cpu"
 # Initialize the model and processor
-model = AutoModelForCausalLM.from_pretrained(args.model_path, torch_dtype=torch_dtype, trust_remote_code=True).to(device)
+model = AutoModelForCausalLM.from_pretrained(args.model_path, trust_remote_code=True).to(device)
 processor = AutoProcessor.from_pretrained(args.model_path, trust_remote_code=True)
 
 lock = threading.Lock()  # Use a lock to ensure thread safety when accessing the model
@@ -38,7 +26,7 @@ def predict_image(image, task: str = "<OD>", prompt: str = None):
     prompt = task + " " + prompt if prompt else task
     print(f"Prompt: {prompt}")
     with lock:
-        inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
+        inputs = processor(text=prompt, images=image, return_tensors="pt").to(device)
         generated_ids = model.generate(
             input_ids=inputs["input_ids"],
             pixel_values=inputs["pixel_values"],