今天小编给大家分享一下怎么实现PyTorch模型编译的相关知识点,内容详细,逻辑清晰,相信大部分人都还太了解这方面的知识,所以分享这篇文章给大家参考一下,希望大家阅读完这篇文章后有所收获,下面我们一起来了解一下吧。
首先应安装 PyTorch。此外,还应安装 TorchVision,并将其作为模型合集 (model zoo)。
可通过 pip 快速安装:
pip install torch==1.7.0 pip install torchvision==0.8.1
或参考官网:pytorch.org/get-started…
PyTorch 版本应该和 TorchVision 版本兼容。
目前 TVM 支持 PyTorch 1.7 和 1.4,其他版本可能不稳定。
import tvm from tvm import relay import numpy as np from tvm.contrib.download import download_testdata # 导入 PyTorch import torch import torchvision
model_name = "resnet18" model = getattr(torchvision.models, model_name)(pretrained=True) model = model.eval() # 通过追踪获取 TorchScripted 模型 input_shape = [1, 3, 224, 224] input_data = torch.randn(input_shape) scripted_model = torch.jit.trace(model, input_data).eval() 输出结果:
Downloading: "download.pytorch.org/models/resn…" to /workspace/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
0%| | 0.00/44.7M [00:00<?, ?B/s] 11%|# | 4.87M/44.7M [00:00<00:00, 51.0MB/s] 22%|##1 | 9.73M/44.7M [00:00<00:00, 49.2MB/s] 74%|#######3 | 32.9M/44.7M [00:00<00:00, 136MB/s] 100%|##########| 44.7M/44.7M [00:00<00:00, 129MB/s]
经典的猫咪示例:
from PIL import Image img_url = "https://cache.yisu.com/upload/information/20230426/112/17141.png?raw=true" img_path = download_testdata(img_url, "cat.png", module="data") img = Image.open(img_path).resize((224, 224)) # 预处理图像,并将其转换为张量 from torchvision import transforms my_preprocess = transforms.Compose( [ transforms.Resize(256), transforms.CenterCrop(224), transforms.ToTensor(), transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), ] ) img = my_preprocess(img) img = np.expand_dims(img, 0)
将 PyTorch 计算图转换为 Relay 计算图。input_name 可以是任意值。
input_name = "input0" shape_list = [(input_name, img.shape)] mod, params = relay.frontend.from_pytorch(scripted_model, shape_list)
用给定的输入规范,将计算图编译为 llvm target。
target = tvm.target.Target("llvm", host="llvm") dev = tvm.cpu(0) with tvm.transform.PassContext(opt_level=3): lib = relay.build(mod, target=target, params=params)
输出结果:
/workspace/python/tvm/driver/build_module.py:268: UserWarning: target_host parameter is going to be deprecated. Please pass in tvm.target.Target(target, host=target_host) instead.
"target_host parameter is going to be deprecated. "
将编译好的模型部署到 target 上:
from tvm.contrib import graph_executor dtype = "float32" m = graph_executor.GraphModule(lib["default"](dev)) # 设置输入 m.set_input(input_name, tvm.nd.array(img.astype(dtype))) # 执行 m.run() # 得到输出 tvm_output = m.get_output(0)
在 1000 个类的分类集中,查找分数最高的第一个:
synset_url = "".join( [ "https://raw.githubusercontent.com/Cadene/", "pretrained-models.pytorch/master/data/", "imagenet_synsets.txt", ] ) synset_name = "imagenet_synsets.txt" synset_path = download_testdata(synset_url, synset_name, module="data") with open(synset_path) as f: synsets = f.readlines() synsets = [x.strip() for x in synsets] splits = [line.split(" ") for line in synsets] key_to_classname = {spl[0]: " ".join(spl[1:]) for spl in splits} class_url = "".join( [ "https://raw.githubusercontent.com/Cadene/", "pretrained-models.pytorch/master/data/", "imagenet_classes.txt", ] ) class_name = "imagenet_classes.txt" class_path = download_testdata(class_url, class_name, module="data") with open(class_path) as f: class_id_to_key = f.readlines() class_id_to_key = [x.strip() for x in class_id_to_key] # 获得 TVM 的前 1 个结果 top1_tvm = np.argmax(tvm_output.numpy()[0]) tvm_class_key = class_id_to_key[top1_tvm] # 将输入转换为 PyTorch 变量,并获取 PyTorch 结果进行比较 with torch.no_grad(): torch_img = torch.from_numpy(img) output = model(torch_img) # 获得 PyTorch 的前 1 个结果 top1_torch = np.argmax(output.numpy()) torch_class_key = class_id_to_key[top1_torch] print("Relay top-1 id: {}, class name: {}".format(top1_tvm, key_to_classname[tvm_class_key])) print("Torch top-1 id: {}, class name: {}".format(top1_torch, key_to_classname[torch_class_key]))
输出结果:
Relay top-1 id: 281, class name: tabby, tabby cat
Torch top-1 id: 281, class name: tabby, tabby cat
以上就是“怎么实现PyTorch模型编译”这篇文章的所有内容,感谢各位的阅读!相信大家阅读完这篇文章都有很大的收获,小编每天都会为大家更新不同的知识,如果还想学习更多的知识,请关注亿速云行业资讯频道。
免责声明:本站发布的内容(图片、视频和文字)以原创、转载和分享为主,文章观点不代表本网站立场,如果涉及侵权请联系站长邮箱:is@yisu.com进行举报,并提供相关证据,一经查实,将立刻删除涉嫌侵权内容。