π [Update] README and the TRT ipynb tutorial
Browse files- README.md +1 -1
- examples/notebook_TensorRT.ipynb +130 -0
README.md
CHANGED
@@ -47,7 +47,7 @@ pip install -r requirements.txt
|
|
47 |
| ------------------ | :---------: | :-------: | :-------: |
|
48 |
| PyTorch | v1.12 | v2.3+ | v1.12 |
|
49 |
| ONNX | β
| β
| - |
|
50 |
-
| TensorRT |
|
51 |
| OpenVINO | - | π§ͺ | β |
|
52 |
|
53 |
</td></tr> </table>
|
|
|
47 |
| ------------------ | :---------: | :-------: | :-------: |
|
48 |
| PyTorch | v1.12 | v2.3+ | v1.12 |
|
49 |
| ONNX | β
| β
| - |
|
50 |
+
| TensorRT | β
| - | - |
|
51 |
| OpenVINO | - | π§ͺ | β |
|
52 |
|
53 |
</td></tr> </table>
|
examples/notebook_TensorRT.ipynb
ADDED
@@ -0,0 +1,130 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "code",
|
5 |
+
"execution_count": null,
|
6 |
+
"metadata": {},
|
7 |
+
"outputs": [],
|
8 |
+
"source": [
|
9 |
+
"import os\n",
|
10 |
+
"import sys\n",
|
11 |
+
"from pathlib import Path\n",
|
12 |
+
"\n",
|
13 |
+
"import torch\n",
|
14 |
+
"from PIL import Image \n",
|
15 |
+
"from loguru import logger\n",
|
16 |
+
"from omegaconf import OmegaConf\n",
|
17 |
+
"\n",
|
18 |
+
"project_root = Path().resolve().parent\n",
|
19 |
+
"sys.path.append(str(project_root))\n",
|
20 |
+
"\n",
|
21 |
+
"from yolo import AugmentationComposer, bbox_nms, create_model, custom_logger, draw_bboxes, Vec2Box\n",
|
22 |
+
"from yolo.config.config import NMSConfig"
|
23 |
+
]
|
24 |
+
},
|
25 |
+
{
|
26 |
+
"cell_type": "code",
|
27 |
+
"execution_count": null,
|
28 |
+
"metadata": {},
|
29 |
+
"outputs": [],
|
30 |
+
"source": [
|
31 |
+
"MODEL = \"v9-c\"\n",
|
32 |
+
"DEVICE = \"cuda:0\"\n",
|
33 |
+
"\n",
|
34 |
+
"WEIGHT_PATH = f\"../weights/{MODEL}.pt\" \n",
|
35 |
+
"TRT_WEIGHT_PATH = f\"../weights/{MODEL}.trt\"\n",
|
36 |
+
"MODEL_CONFIG = f\"../yolo/config/model/{MODEL}.yaml\"\n",
|
37 |
+
"\n",
|
38 |
+
"IMAGE_PATH = \"../demo/images/inference/image.png\"\n",
|
39 |
+
"IMAGE_SIZE = (640, 640)\n",
|
40 |
+
"\n",
|
41 |
+
"custom_logger()\n",
|
42 |
+
"device = torch.device(DEVICE)\n",
|
43 |
+
"image = Image.open(IMAGE_PATH)"
|
44 |
+
]
|
45 |
+
},
|
46 |
+
{
|
47 |
+
"cell_type": "code",
|
48 |
+
"execution_count": null,
|
49 |
+
"metadata": {},
|
50 |
+
"outputs": [],
|
51 |
+
"source": [
|
52 |
+
"if os.path.exists(TRT_WEIGHT_PATH):\n",
|
53 |
+
" from torch2trt import TRTModule\n",
|
54 |
+
"\n",
|
55 |
+
" model_trt = TRTModule()\n",
|
56 |
+
" model_trt.load_state_dict(torch.load(TRT_WEIGHT_PATH))\n",
|
57 |
+
"else:\n",
|
58 |
+
" from torch2trt import torch2trt\n",
|
59 |
+
"\n",
|
60 |
+
" with open(MODEL_CONFIG) as stream:\n",
|
61 |
+
" cfg_model = OmegaConf.load(stream)\n",
|
62 |
+
"\n",
|
63 |
+
" model = create_model(cfg_model, weight_path=WEIGHT_PATH)\n",
|
64 |
+
" model = model.to(device).eval()\n",
|
65 |
+
"\n",
|
66 |
+
" dummy_input = torch.ones((1, 3, 640, 640)).to(device)\n",
|
67 |
+
" logger.info(f\"β»οΈ Creating TensorRT model\")\n",
|
68 |
+
" model_trt = torch2trt(model, [dummy_input])\n",
|
69 |
+
" torch.save(model_trt.state_dict(), TRT_WEIGHT_PATH)\n",
|
70 |
+
" logger.info(f\"π₯ TensorRT model saved to oonx.pt\")\n",
|
71 |
+
"\n",
|
72 |
+
"transform = AugmentationComposer([], IMAGE_SIZE)\n",
|
73 |
+
"vec2box = Vec2Box(model_trt, IMAGE_SIZE, device)\n"
|
74 |
+
]
|
75 |
+
},
|
76 |
+
{
|
77 |
+
"cell_type": "code",
|
78 |
+
"execution_count": null,
|
79 |
+
"metadata": {},
|
80 |
+
"outputs": [],
|
81 |
+
"source": [
|
82 |
+
"image, bbox = transform(image, torch.zeros(0, 5))\n",
|
83 |
+
"image = image.to(device)[None]"
|
84 |
+
]
|
85 |
+
},
|
86 |
+
{
|
87 |
+
"cell_type": "code",
|
88 |
+
"execution_count": null,
|
89 |
+
"metadata": {},
|
90 |
+
"outputs": [],
|
91 |
+
"source": [
|
92 |
+
"with torch.no_grad():\n",
|
93 |
+
" predict = model_trt(image)\n",
|
94 |
+
" predict = vec2box(predict[\"Main\"])\n",
|
95 |
+
"predict_box = bbox_nms(predict[0], predict[2], NMSConfig(0.5, 0.5))\n",
|
96 |
+
"draw_bboxes(image, predict_box)"
|
97 |
+
]
|
98 |
+
},
|
99 |
+
{
|
100 |
+
"cell_type": "markdown",
|
101 |
+
"metadata": {},
|
102 |
+
"source": [
|
103 |
+
"Sample Output:\n",
|
104 |
+
"\n",
|
105 |
+
""
|
106 |
+
]
|
107 |
+
}
|
108 |
+
],
|
109 |
+
"metadata": {
|
110 |
+
"kernelspec": {
|
111 |
+
"display_name": "yolomit",
|
112 |
+
"language": "python",
|
113 |
+
"name": "python3"
|
114 |
+
},
|
115 |
+
"language_info": {
|
116 |
+
"codemirror_mode": {
|
117 |
+
"name": "ipython",
|
118 |
+
"version": 3
|
119 |
+
},
|
120 |
+
"file_extension": ".py",
|
121 |
+
"mimetype": "text/x-python",
|
122 |
+
"name": "python",
|
123 |
+
"nbconvert_exporter": "python",
|
124 |
+
"pygments_lexer": "ipython3",
|
125 |
+
"version": "3.1.undefined"
|
126 |
+
}
|
127 |
+
},
|
128 |
+
"nbformat": 4,
|
129 |
+
"nbformat_minor": 2
|
130 |
+
}
|