Skip to content

Commit c5e9cd2

Browse files
ZwwWayneZwwWayne
committed
[Refactor]: refactor configs of FP16 models (open-mmlab#6592)
* refactor configs of fp16 * update * update * update
1 parent 5188920 commit c5e9cd2

18 files changed

+184
-199
lines changed

‎configs/dcn/README.md

+2
Original file line numberDiff line numberDiff line change
@@ -43,6 +43,8 @@
4343
| R-50-FPN | Cascade Mask | pytorch | dconv(c3-c5) | - | 1x | 6.0 | 10.0 | 44.4 | 38.6 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/dcn/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco_20200202-42e767a2.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco_20200202_010309.log.json) |
4444
| R-101-FPN | Cascade Mask | pytorch | dconv(c3-c5) | - | 1x | 8.0 | 8.6 | 45.8 | 39.7 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/dcn/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco_20200204-df0c5f10.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco_20200204_134006.log.json) |
4545
| X-101-32x4d-FPN | Cascade Mask | pytorch | dconv(c3-c5) | - | 1x | 9.2 | | 47.3 | 41.1 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/dcn/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco-e75f90c8.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco-20200606_183737.log.json) |
46+
| R-50-FPN (FP16) | Mask | pytorch | dconv(c3-c5) | - | 1x | 3.0 | | 41.9 | 37.5 |[config](https://github.com/open-mmlab/mmdetection/tree/master/configs/fp16/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco_20210520_180247-c06429d2.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco_20210520_180247.log.json) |
47+
| R-50-FPN (FP16) | Mask | pytorch | mdconv(c3-c5)| - | 1x | 3.1 | | 42.0 | 37.6 |[config](https://github.com/open-mmlab/mmdetection/tree/master/configs/fp16/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco_20210520_180434-cf8fefa5.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco_20210520_180434.log.json) |
4648

4749
**Notes:**
4850

‎configs/fp16/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco.py ‎configs/dcn/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco.py

File renamed without changes.

‎configs/fp16/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco.py ‎configs/dcn/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco.py

File renamed without changes.

‎configs/dcn/metafile.yml

+49-7
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@ Collections:
1010
- Deformable Convolution
1111
Paper:
1212
URL: https://arxiv.org/abs/1811.11168
13-
Title: 'Deformable ConvNets v2: More Deformable, Better Results'
13+
Title: "Deformable ConvNets v2: More Deformable, Better Results"
1414
README: configs/dcn/README.md
1515
Code:
1616
URL: https://github.com/open-mmlab/mmdetection/blob/v2.0.0/mmdet/ops/dcn/deform_conv.py#L15
@@ -178,7 +178,7 @@ Models:
178178
- Task: Instance Segmentation
179179
Dataset: COCO
180180
Metrics:
181-
mask AP: 37.4
181+
mask AP: 37.4
182182
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco_20200203-4d9ad43b.pth
183183

184184
- Name: mask_rcnn_r50_fpn_mdconv_c3-c5_1x_coco
@@ -202,9 +202,51 @@ Models:
202202
- Task: Instance Segmentation
203203
Dataset: COCO
204204
Metrics:
205-
mask AP: 37.1
205+
mask AP: 37.1
206206
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/mask_rcnn_r50_fpn_mdconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_mdconv_c3-c5_1x_coco_20200203-ad97591f.pth
207207

208+
- Name: mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco
209+
In Collection: Deformable Convolutional Networks
210+
Config: configs/dcn/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco.py
211+
Metadata:
212+
Training Techniques:
213+
- SGD with Momentum
214+
- Weight Decay
215+
- Mixed Precision Training
216+
Training Memory (GB): 3.0
217+
Epochs: 12
218+
Results:
219+
- Task: Object Detection
220+
Dataset: COCO
221+
Metrics:
222+
box AP: 41.9
223+
- Task: Instance Segmentation
224+
Dataset: COCO
225+
Metrics:
226+
mask AP: 37.5
227+
Weights: https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_dconv_c3-c5_1x_coco_20210520_180247-c06429d2.pth
228+
229+
- Name: mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco
230+
In Collection: Deformable Convolutional Networks
231+
Config: configs/dcn/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco.py
232+
Metadata:
233+
Training Memory (GB): 3.1
234+
Training Techniques:
235+
- SGD with Momentum
236+
- Weight Decay
237+
- Mixed Precision Training
238+
Epochs: 12
239+
Results:
240+
- Task: Object Detection
241+
Dataset: COCO
242+
Metrics:
243+
box AP: 42.0
244+
- Task: Instance Segmentation
245+
Dataset: COCO
246+
Metrics:
247+
mask AP: 37.6
248+
Weights: https://download.openmmlab.com/mmdetection/v2.0/fp16/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco/mask_rcnn_r50_fpn_fp16_mdconv_c3-c5_1x_coco_20210520_180434-cf8fefa5.pth
249+
208250
- Name: mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco
209251
In Collection: Deformable Convolutional Networks
210252
Config: configs/dcn/mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco.py
@@ -226,7 +268,7 @@ Models:
226268
- Task: Instance Segmentation
227269
Dataset: COCO
228270
Metrics:
229-
mask AP: 38.9
271+
mask AP: 38.9
230272
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco/mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco_20200216-a71f5bce.pth
231273

232274
- Name: cascade_rcnn_r50_fpn_dconv_c3-c5_1x_coco
@@ -290,7 +332,7 @@ Models:
290332
- Task: Instance Segmentation
291333
Dataset: COCO
292334
Metrics:
293-
mask AP: 38.6
335+
mask AP: 38.6
294336
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r50_fpn_dconv_c3-c5_1x_coco_20200202-42e767a2.pth
295337

296338
- Name: cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco
@@ -314,7 +356,7 @@ Models:
314356
- Task: Instance Segmentation
315357
Dataset: COCO
316358
Metrics:
317-
mask AP: 39.7
359+
mask AP: 39.7
318360
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_r101_fpn_dconv_c3-c5_1x_coco_20200204-df0c5f10.pth
319361

320362
- Name: cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco
@@ -331,5 +373,5 @@ Models:
331373
- Task: Instance Segmentation
332374
Dataset: COCO
333375
Metrics:
334-
mask AP: 41.1
376+
mask AP: 41.1
335377
Weights: https://download.openmmlab.com/mmdetection/v2.0/dcn/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco/cascade_mask_rcnn_x101_32x4d_fpn_dconv_c3-c5_1x_coco-e75f90c8.pth

‎configs/faster_rcnn/README.md

+1
Original file line numberDiff line numberDiff line change
@@ -22,6 +22,7 @@
2222
| R-50-DC5 | caffe | 1x | - | - | 37.2 | [config](https://github.com/open-mmlab/mmdetection/blob/master/configs/faster_rcnn/faster_rcnn_r50_caffe_dc5_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_caffe_dc5_1x_coco/faster_rcnn_r50_caffe_dc5_1x_coco_20201030_151909-531f0f43.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_caffe_dc5_1x_coco/faster_rcnn_r50_caffe_dc5_1x_coco_20201030_151909.log.json) |
2323
| R-50-FPN | caffe | 1x | 3.8 | | 37.8 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/faster_rcnn/faster_rcnn_r50_caffe_fpn_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_caffe_fpn_1x_coco/faster_rcnn_r50_caffe_fpn_1x_coco_bbox_mAP-0.378_20200504_180032-c5925ee5.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_caffe_fpn_1x_coco/faster_rcnn_r50_caffe_fpn_1x_coco_20200504_180032.log.json) |
2424
| R-50-FPN | pytorch | 1x | 4.0 | 21.4 | 37.4 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130_204655.log.json) |
25+
| R-50-FPN (FP16) | pytorch | 1x | 3.4 | 28.8 | 37.5 |[config](https://github.com/open-mmlab/mmdetection/tree/master/configs/fp16/faster_rcnn_r50_fpn_fp16_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/fp16/faster_rcnn_r50_fpn_fp16_1x_coco/faster_rcnn_r50_fpn_fp16_1x_coco_20200204-d4dc1471.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/fp16/faster_rcnn_r50_fpn_fp16_1x_coco/faster_rcnn_r50_fpn_fp16_1x_coco_20200204_143530.log.json) |
2526
| R-50-FPN | pytorch | 2x | - | - | 38.4 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/faster_rcnn/faster_rcnn_r50_fpn_2x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_2x_coco/faster_rcnn_r50_fpn_2x_coco_bbox_mAP-0.384_20200504_210434-a5d8aa15.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_2x_coco/faster_rcnn_r50_fpn_2x_coco_20200504_210434.log.json) |
2627
| R-101-FPN | caffe | 1x | 5.7 | | 39.8 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/faster_rcnn/faster_rcnn_r101_caffe_fpn_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r101_caffe_fpn_1x_coco/faster_rcnn_r101_caffe_fpn_1x_coco_bbox_mAP-0.398_20200504_180057-b269e9dd.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r101_caffe_fpn_1x_coco/faster_rcnn_r101_caffe_fpn_1x_coco_20200504_180057.log.json) |
2728
| R-101-FPN | pytorch | 1x | 6.0 | 15.6 | 39.4 | [config](https://github.com/open-mmlab/mmdetection/tree/master/configs/faster_rcnn/faster_rcnn_r101_fpn_1x_coco.py) | [model](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r101_fpn_1x_coco/faster_rcnn_r101_fpn_1x_coco_20200130-f513f705.pth) | [log](https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r101_fpn_1x_coco/faster_rcnn_r101_fpn_1x_coco_20200130_204655.log.json) |

‎configs/faster_rcnn/faster_rcnn_r50_fpn_fp16_1x_coco.py

+3
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
_base_ = './faster_rcnn_r50_fpn_1x_coco.py'
2+
# fp16 settings
3+
fp16 = dict(loss_scale=512.)

‎configs/faster_rcnn/metafile.yml

+67-43
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,7 @@ Collections:
1313
- RoIPool
1414
Paper:
1515
URL: https://arxiv.org/abs/1506.01497
16-
Title: 'Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks'
16+
Title: "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks"
1717
README: configs/faster_rcnn/README.md
1818
Code:
1919
URL: https://github.com/open-mmlab/mmdetection/blob/v2.0.0/mmdet/models/detectors/faster_rcnn.py#L6
@@ -65,18 +65,42 @@ Models:
6565
box AP: 37.4
6666
Weights: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth
6767

68+
- Name: faster_rcnn_r50_fpn_fp16_1x_coco
69+
In Collection: Faster R-CNN
70+
Config: configs/faster_rcnn/faster_rcnn_r50_fpn_fp16_1x_coco.py
71+
Metadata:
72+
Training Memory (GB): 3.4
73+
Training Techniques:
74+
- SGD with Momentum
75+
- Weight Decay
76+
- Mixed Precision Training
77+
inference time (ms/im):
78+
- value: 34.72
79+
hardware: V100
80+
backend: PyTorch
81+
batch size: 1
82+
mode: FP16
83+
resolution: (800, 1333)
84+
Epochs: 12
85+
Results:
86+
- Task: Object Detection
87+
Dataset: COCO
88+
Metrics:
89+
box AP: 37.5
90+
Weights: https://download.openmmlab.com/mmdetection/v2.0/fp16/faster_rcnn_r50_fpn_fp16_1x_coco/faster_rcnn_r50_fpn_fp16_1x_coco_20200204-d4dc1471.pth
91+
6892
- Name: faster_rcnn_r50_fpn_2x_coco
6993
In Collection: Faster R-CNN
7094
Config: configs/faster_rcnn/faster_rcnn_r50_fpn_2x_coco.py
7195
Metadata:
7296
Training Memory (GB): 4.0
7397
inference time (ms/im):
74-
- value: 46.73
75-
hardware: V100
76-
backend: PyTorch
77-
batch size: 1
78-
mode: FP32
79-
resolution: (800, 1333)
98+
- value: 46.73
99+
hardware: V100
100+
backend: PyTorch
101+
batch size: 1
102+
mode: FP32
103+
resolution: (800, 1333)
80104
Epochs: 24
81105
Results:
82106
- Task: Object Detection
@@ -104,12 +128,12 @@ Models:
104128
Metadata:
105129
Training Memory (GB): 6.0
106130
inference time (ms/im):
107-
- value: 64.1
108-
hardware: V100
109-
backend: PyTorch
110-
batch size: 1
111-
mode: FP32
112-
resolution: (800, 1333)
131+
- value: 64.1
132+
hardware: V100
133+
backend: PyTorch
134+
batch size: 1
135+
mode: FP32
136+
resolution: (800, 1333)
113137
Epochs: 12
114138
Results:
115139
- Task: Object Detection
@@ -124,12 +148,12 @@ Models:
124148
Metadata:
125149
Training Memory (GB): 6.0
126150
inference time (ms/im):
127-
- value: 64.1
128-
hardware: V100
129-
backend: PyTorch
130-
batch size: 1
131-
mode: FP32
132-
resolution: (800, 1333)
151+
- value: 64.1
152+
hardware: V100
153+
backend: PyTorch
154+
batch size: 1
155+
mode: FP32
156+
resolution: (800, 1333)
133157
Epochs: 24
134158
Results:
135159
- Task: Object Detection
@@ -144,12 +168,12 @@ Models:
144168
Metadata:
145169
Training Memory (GB): 7.2
146170
inference time (ms/im):
147-
- value: 72.46
148-
hardware: V100
149-
backend: PyTorch
150-
batch size: 1
151-
mode: FP32
152-
resolution: (800, 1333)
171+
- value: 72.46
172+
hardware: V100
173+
backend: PyTorch
174+
batch size: 1
175+
mode: FP32
176+
resolution: (800, 1333)
153177
Epochs: 12
154178
Results:
155179
- Task: Object Detection
@@ -164,12 +188,12 @@ Models:
164188
Metadata:
165189
Training Memory (GB): 7.2
166190
inference time (ms/im):
167-
- value: 72.46
168-
hardware: V100
169-
backend: PyTorch
170-
batch size: 1
171-
mode: FP32
172-
resolution: (800, 1333)
191+
- value: 72.46
192+
hardware: V100
193+
backend: PyTorch
194+
batch size: 1
195+
mode: FP32
196+
resolution: (800, 1333)
173197
Epochs: 24
174198
Results:
175199
- Task: Object Detection
@@ -184,12 +208,12 @@ Models:
184208
Metadata:
185209
Training Memory (GB): 10.3
186210
inference time (ms/im):
187-
- value: 106.38
188-
hardware: V100
189-
backend: PyTorch
190-
batch size: 1
191-
mode: FP32
192-
resolution: (800, 1333)
211+
- value: 106.38
212+
hardware: V100
213+
backend: PyTorch
214+
batch size: 1
215+
mode: FP32
216+
resolution: (800, 1333)
193217
Epochs: 12
194218
Results:
195219
- Task: Object Detection
@@ -204,12 +228,12 @@ Models:
204228
Metadata:
205229
Training Memory (GB): 10.3
206230
inference time (ms/im):
207-
- value: 106.38
208-
hardware: V100
209-
backend: PyTorch
210-
batch size: 1
211-
mode: FP32
212-
resolution: (800, 1333)
231+
- value: 106.38
232+
hardware: V100
233+
backend: PyTorch
234+
batch size: 1
235+
mode: FP32
236+
resolution: (800, 1333)
213237
Epochs: 24
214238
Results:
215239
- Task: Object Detection

‎configs/fp16/README.md

-24
This file was deleted.

‎configs/fp16/faster_rcnn_r50_fpn_fp16_1x_coco.py

-3
This file was deleted.

‎configs/fp16/mask_rcnn_r50_fpn_fp16_1x_coco.py

-3
This file was deleted.

0 commit comments

Comments
 (0)