Update onnx/builder.py
Browse files- onnx/builder.py +4 -4
onnx/builder.py
CHANGED
|
@@ -17,7 +17,7 @@ def build_vision(args):
|
|
| 17 |
prompt = f"{user_prompt}<|image_1|>\nWhat is shown in this image?{prompt_suffix}{assistant_prompt}"
|
| 18 |
url = "https://www.ilankelman.org/stopsigns/australia.jpg"
|
| 19 |
image = Image.open(requests.get(url, stream=True).raw)
|
| 20 |
-
inputs = processor(prompt, image, return_tensors="pt").to(args.execution_provider)
|
| 21 |
inputs["pixel_values"] = inputs["pixel_values"].to(args.precision)
|
| 22 |
|
| 23 |
# TorchScript export
|
|
@@ -214,8 +214,8 @@ def get_args():
|
|
| 214 |
"-e",
|
| 215 |
"--execution_provider",
|
| 216 |
required=True,
|
| 217 |
-
choices=["cpu", "cuda"],
|
| 218 |
-
help="
|
| 219 |
)
|
| 220 |
|
| 221 |
parser.add_argument(
|
|
@@ -238,7 +238,7 @@ if __name__ == "__main__":
|
|
| 238 |
args = get_args()
|
| 239 |
config = AutoConfig.from_pretrained(args.input, trust_remote_code=True)
|
| 240 |
processor = AutoProcessor.from_pretrained(args.input, trust_remote_code=True)
|
| 241 |
-
model = AutoModelForCausalLM.from_pretrained(args.input, trust_remote_code=True, torch_dtype=args.precision).to(args.execution_provider)
|
| 242 |
|
| 243 |
# Build model components
|
| 244 |
build_vision(args)
|
|
|
|
| 17 |
prompt = f"{user_prompt}<|image_1|>\nWhat is shown in this image?{prompt_suffix}{assistant_prompt}"
|
| 18 |
url = "https://www.ilankelman.org/stopsigns/australia.jpg"
|
| 19 |
image = Image.open(requests.get(url, stream=True).raw)
|
| 20 |
+
inputs = processor(prompt, image, return_tensors="pt").to(args.execution_provider.replace("dml", "cuda"))
|
| 21 |
inputs["pixel_values"] = inputs["pixel_values"].to(args.precision)
|
| 22 |
|
| 23 |
# TorchScript export
|
|
|
|
| 214 |
"-e",
|
| 215 |
"--execution_provider",
|
| 216 |
required=True,
|
| 217 |
+
choices=["cpu", "cuda", "dml"],
|
| 218 |
+
help="Execution provider for Phi-3 vision components",
|
| 219 |
)
|
| 220 |
|
| 221 |
parser.add_argument(
|
|
|
|
| 238 |
args = get_args()
|
| 239 |
config = AutoConfig.from_pretrained(args.input, trust_remote_code=True)
|
| 240 |
processor = AutoProcessor.from_pretrained(args.input, trust_remote_code=True)
|
| 241 |
+
model = AutoModelForCausalLM.from_pretrained(args.input, trust_remote_code=True, torch_dtype=args.precision).to(args.execution_provider.replace("dml", "cuda"))
|
| 242 |
|
| 243 |
# Build model components
|
| 244 |
build_vision(args)
|