File size: 1,715 Bytes
59042ec
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
import json
import spu.utils.distributed as ppd

from time import time
from datasets import load_dataset
from transformers import (
    AutoConfig,
    AutoImageProcessor,
    FlaxResNetForImageClassification,
)

parser = argparse.ArgumentParser(description='distributed driver.')
parser.add_argument("-c", "--config", default="3pc.json")
args = parser.parse_args()

with open(args.config, 'r') as file:
    conf = json.load(file)

ppd.init(conf["nodes"], conf["devices"])

dataset = load_dataset("huggingface/cats-image")
image = dataset["test"]["image"][0]

processor = AutoImageProcessor.from_pretrained("microsoft/resnet-50")
model = FlaxResNetForImageClassification.from_pretrained("microsoft/resnet-50")

inputs = processor(image, return_tensors="jax")["pixel_values"]


def run_on_spu(inputs, model):
    start = time()
    inputs = ppd.device("P1")(lambda x: x)(inputs)
    params = ppd.device("P2")(lambda x: x)(model.params)
    outputs = ppd.device("SPU")(inference)(inputs, params)
    outputs = ppd.get(outputs)
    outputs = outputs['logits']
    predicted_class_idx = jax.numpy.argmax(outputs, axis=-1)
    print(f"Elapsed time:{time() - start}")
    print("Predicted class:", model.config.id2label[predicted_class_idx.item()])


def run_on_cpu(inputs, model):
    start = time()
    outputs = inference(inputs, model.params)
    outputs = outputs['logits']
    predicted_class_idx = jax.numpy.argmax(outputs, axis=-1)
    print(f"Elapsed time:{time() - start}")
    print("Predicted class:", model.config.id2label[predicted_class_idx.item()])


if __name__ == "__main__":
    print("Run on CPU\n------\n")
    run_on_cpu(inputs, model)
    print("Run on SPU\n------\n")
    run_on_spu(inputs, model)