[Fix] Remove `fp16` folder in `configs` (#1074)

* remove fp16 folders

* fix metafile.yml

* fix typos

* fix model_zoo.md

* unify collections

* fix typos

* fix_typos

.dev_scripts/test_benchmark.sh

@@ -25,11 +25,11 @@ GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_test.sh $PARTITION fcos3d_r
 $CHECKPOINT_DIR/configs/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d.py/latest.pth --eval map \
 2>&1|tee $CHECKPOINT_DIR/configs/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d.py/FULL_LOG.txt &
 
-echo 'configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py' &
-mkdir -p $CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
-GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_test.sh $PARTITION hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py \
-$CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/latest.pth --eval map \
-2>&1|tee $CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/FULL_LOG.txt &
+echo 'configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py' &
+mkdir -p $CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
+GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_test.sh $PARTITION hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py \
+$CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/latest.pth --eval map \
+2>&1|tee $CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/FULL_LOG.txt &
 
 echo 'configs/free_anchor/hv_pointpillars_regnet-1.6gf_fpn_sbn-all_free-anchor_strong-aug_4x8_3x_nus-3d.py' &
 mkdir -p $CHECKPOINT_DIR/configs/free_anchor/hv_pointpillars_regnet-1.6gf_fpn_sbn-all_free-anchor_strong-aug_4x8_3x_nus-3d.py

.dev_scripts/train_benchmark.sh

@@ -25,11 +25,11 @@ GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_train.sh $PARTITION fcos3d_
 $CHECKPOINT_DIR/configs/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d.py --cfg-options checkpoint_config.max_keep_ckpts=1 \
 2>&1|tee $CHECKPOINT_DIR/configs/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d.py/FULL_LOG.txt &
 
-echo 'configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py' &
-mkdir -p $CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
-GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_train.sh $PARTITION hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py \
-$CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py --cfg-options checkpoint_config.max_keep_ckpts=1 \
-2>&1|tee $CHECKPOINT_DIR/configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/FULL_LOG.txt &
+echo 'configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py' &
+mkdir -p $CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
+GPUS=8 GPUS_PER_NODE=8 CPUS_PER_TASK=5 ./tools/slurm_train.sh $PARTITION hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py \
+$CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py --cfg-options checkpoint_config.max_keep_ckpts=1 \
+2>&1|tee $CHECKPOINT_DIR/configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py/FULL_LOG.txt &
 
 echo 'configs/free_anchor/hv_pointpillars_regnet-1.6gf_fpn_sbn-all_free-anchor_strong-aug_4x8_3x_nus-3d.py' &
 mkdir -p $CHECKPOINT_DIR/configs/free_anchor/hv_pointpillars_regnet-1.6gf_fpn_sbn-all_free-anchor_strong-aug_4x8_3x_nus-3d.py

configs/fp16/README.md

-# Mixed Precision Training
-
-## Introduction
-
-<!-- [OTHERS] -->
-
-We implement mixed precision training and apply it to VoxelNets (e.g., SECOND and PointPillars).
-The results are in the following tables.
-
-**Note**: For mixed precision training, we currently do not support PointNet-based methods (e.g., VoteNet).
-Mixed precision training for PointNet-based methods will be supported in the future release.
-
-## Results
-
-### SECOND on KITTI dataset
-
-|  Backbone   |Class| Lr schd | FP32 Mem (GB) | FP16 Mem (GB) | FP32 mAP | FP16 mAP |Download |
-| :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: | :------: |
-|    [SECFPN](./hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py)| Car |cyclic 80e|5.4|2.9|79.07|78.72|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301-1f5ad833.pth)&#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301.log.json)|
-|    [SECFPN](./hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py)| 3 Class |cyclic 80e|5.4|2.9|64.41|67.4|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059-05f67bdf.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059.log.json)|
-
-### PointPillars on nuScenes dataset
-
-|  Backbone   | Lr schd | FP32 Mem (GB) | FP16 Mem (GB) | FP32 mAP | FP32 NDS| FP16 mAP | FP16 NDS| Download |
-| :---------: | :-----: | :------: | :------------: | :----: |:----: | :----: |:----: | :------: |
-|[SECFPN](./hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py)|2x|16.4|8.37|35.17|49.7|35.19|50.27|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626-c3f0483e.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626.log.json)|
-|[FPN](./hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py)|2x|16.4|8.40|40.0|53.3|39.26|53.26|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719-269f9dd6.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719.log.json)|
-
-**Note**:
-1. With mixed precision training, we can train PointPillars with nuScenes dataset on 8 Titan XP GPUS with batch size of 2.
-This will cause OOM error without mixed precision training.
-2. The loss scale for PointPillars on nuScenes dataset is specifically tuned to avoid the loss to be Nan. We find 32 is more stable than 512, though loss scale 32 still cause Nan sometimes.

configs/fp16/metafile.yml

-Collections:
-  - Name: FP16
-    Metadata:
-      Training Techniques:
-        - AdamW
-        - Mixed Precision Training
-      Training Resources: 8x TITAN Xp
-      Architecture:
-        - Hard Voxelization
-    Paper:
-      URL: https://arxiv.org/abs/1710.03740
-      Title: 'Mixed Precision Training'
-    README: configs/fp16/README.md
-    Code:
-      Version: v0.7.0
-
-Models:
-  - Name: hv_second_secfpn_fp16_6x8_80e_kitti-3d-car
-    In Collection: FP16
-    Config: configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py
-    Metadata:
-      Training Data: KITTI
-      Training Memory (GB): 2.9
-    Results:
-      - Task: 3D Object Detection
-        Dataset: KITTI
-        Metrics:
-          FP16 mAP: 78.72
-    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301-1f5ad833.pth
-
-  - Name: hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class
-    In Collection: FP16
-    Config: configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
-    Metadata:
-      Training Data: KITTI
-      Training Memory (GB): 2.9
-    Results:
-      - Task: 3D Object Detection
-        Dataset: KITTI
-        Metrics:
-          FP16 mAP: 67.4
-    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059-05f67bdf.pth
-
-  - Name: hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d
-    In Collection: FP16
-    Config: configs/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py
-    Metadata:
-      Training Data: nuScenes
-      Training Memory (GB): 8.37
-    Results:
-      - Task: 3D Object Detection
-        Dataset: nuScenes
-        Metrics:
-          FP16 mAP: 35.19
-          FP16 NDS: 50.27
-    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626-c3f0483e.pth
-
-  - Name: hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d
-    In Collection: FP16
-    Config: configs/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py
-    Metadata:
-      Training Data: nuScenes
-      Training Memory (GB): 8.40
-    Results:
-      - Task: 3D Object Detection
-        Dataset: nuScenes
-        Metrics:
-          FP16 mAP: 39.26
-          FP16 NDS: 53.26
-    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719-269f9dd6.pth

configs/pointpillars/README.md

@@ -31,7 +31,9 @@ We implement PointPillars and provide the results and checkpoints on KITTI, nuSc
 |  Backbone   | Lr schd | Mem (GB) | Inf time (fps) | mAP |NDS| Download |
 | :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
 |[SECFPN](./hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d.py)|2x|16.4||35.17|49.7|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/pointpillars/hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d_20200620_230725-0817d270.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/pointpillars/hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d_20200620_230725.log.json)|
+|[SECFPN (FP16)](./hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py)|2x|8.37||35.19|50.27|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626-c3f0483e.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626.log.json)|
 |[FPN](./hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d.py)|2x|16.4||40.0|53.3|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/pointpillars/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d_20200620_230405-2fa62f3d.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/pointpillars/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d_20200620_230405.log.json)|
+|[FPN (FP16)](./hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py)|2x|8.40||39.26|53.26|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719-269f9dd6.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719.log.json)|
 
 ### Lyft
 
@@ -62,3 +64,4 @@ We implement PointPillars and provide the results and checkpoints on KITTI, nuSc
 - **Implementation Details**: We basically follow the implementation in the [paper](https://arxiv.org/pdf/1912.04838.pdf) in terms of the network architecture (having a
 stride of 1 for the first convolutional block). Different settings of voxelization, data augmentation and hyper parameters make these baselines outperform those in the paper by about 7 mAP for car and 4 mAP for pedestrian with only a subset of the whole dataset. All of these results are achieved without bells-and-whistles, e.g. ensemble, multi-scale training and test augmentation.
 - **License Aggrement**: To comply the [license agreement of Waymo dataset](https://waymo.com/open/terms/), the pre-trained models on Waymo dataset are not released. We still release the training log as a reference to ease the future research.
+- `FP16` means Mixed Precision (FP16) is adopted in training. With mixed precision training, we can train PointPillars with nuScenes dataset on 8 Titan XP GPUS with batch size of 2. This will cause OOM error without mixed precision training. The loss scale for PointPillars on nuScenes dataset is specifically tuned to avoid the loss to be Nan. We find 32 is more stable than 512, though loss scale 32 still cause Nan sometimes.

configs/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py → configs/pointpillars/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py

-_base_ = '../pointpillars/hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d.py'
+_base_ = './hv_pointpillars_fpn_sbn-all_4x8_2x_nus-3d.py'
 data = dict(samples_per_gpu=2, workers_per_gpu=2)
 # fp16 settings, the loss scale is specifically tuned to avoid Nan
 fp16 = dict(loss_scale=32.)

configs/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py → configs/pointpillars/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py

-_base_ = '../pointpillars/hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d.py'
+_base_ = './hv_pointpillars_secfpn_sbn-all_4x8_2x_nus-3d.py'
 data = dict(samples_per_gpu=2, workers_per_gpu=2)
 # fp16 settings, the loss scale is specifically tuned to avoid Nan
 fp16 = dict(loss_scale=32.)

configs/pointpillars/metafile.yml

@@ -167,3 +167,47 @@ Models:
           mAPH@L1: 63.3
           mAP@L2: 62.6
           mAPH@L2: 57.6
+
+  - Name: hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d
+    In Collection: PointPillars
+    Config: configs/pointpillars/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d.py
+    Metadata:
+      Training Techniques:
+        - AdamW
+        - Mixed Precision Training
+      Training Resources: 8x TITAN Xp
+      Architecture:
+        - Hard Voxelization
+      Training Data: nuScenes
+      Training Memory (GB): 8.37
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 35.19
+          NDS: 50.27
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_secfpn_sbn-all_fp16_2x8_2x_nus-3d_20201020_222626-c3f0483e.pth
+    Code:
+      Version: v0.7.0
+
+  - Name: hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d
+    In Collection: PointPillars
+    Config: configs/pointpillars/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py
+    Metadata:
+      Training Techniques:
+        - AdamW
+        - Mixed Precision Training
+      Training Resources: 8x TITAN Xp
+      Architecture:
+        - Hard Voxelization
+      Training Data: nuScenes
+      Training Memory (GB): 8.40
+    Results:
+      - Task: 3D Object Detection
+        Dataset: nuScenes
+        Metrics:
+          mAP: 39.26
+          NDS: 53.26
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d_20201021_120719-269f9dd6.pth
+    Code:
+      Version: v0.7.0

configs/fp16/hv_pointpillars_regnet-400mf_fpn_sbn-all_fp16_2x8_2x_nus-3d.py → configs/regnet/hv_pointpillars_regnet-400mf_fpn_sbn-all_fp16_2x8_2x_nus-3d.py

-_base_ = '../regnet/hv_pointpillars_regnet-400mf_fpn_sbn-all_4x8_2x_nus-3d.py'
+_base_ = './hv_pointpillars_regnet-400mf_fpn_sbn-all_4x8_2x_nus-3d.py'
 data = dict(samples_per_gpu=2, workers_per_gpu=2)
 # fp16 settings, the loss scale is specifically tuned to avoid Nan
 fp16 = dict(loss_scale=32.)

configs/second/README.md

@@ -23,7 +23,9 @@ We implement SECOND and provide the results and checkpoints on KITTI dataset.
 |  Backbone   |Class| Lr schd | Mem (GB) | Inf time (fps) | mAP |Download |
 | :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
 |    [SECFPN](./hv_second_secfpn_6x8_80e_kitti-3d-car.py)| Car |cyclic 80e|5.4||79.07|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/second/hv_second_secfpn_6x8_80e_kitti-3d-car/hv_second_secfpn_6x8_80e_kitti-3d-car_20200620_230238-393f000c.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/second/hv_second_secfpn_6x8_80e_kitti-3d-car/hv_second_secfpn_6x8_80e_kitti-3d-car_20200620_230238.log.json)|
+|    [SECFPN (FP16)](./hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py)| Car |cyclic 80e|2.9||78.72|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301-1f5ad833.pth)| [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301.log.json)|
 |    [SECFPN](./hv_second_secfpn_6x8_80e_kitti-3d-3class.py)| 3 Class |cyclic 80e|5.4||64.41|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/second/hv_second_secfpn_6x8_80e_kitti-3d-3class/hv_second_secfpn_6x8_80e_kitti-3d-3class_20200620_230238-9208083a.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/second/hv_second_secfpn_6x8_80e_kitti-3d-3class/hv_second_secfpn_6x8_80e_kitti-3d-3class_20200620_230238.log.json)|
+|    [SECFPN (FP16)](./hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py)| 3 Class |cyclic 80e|2.9||67.4|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059-05f67bdf.pth) | [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059.log.json)|
 
 ### Waymo
 
@@ -34,4 +36,7 @@ We implement SECOND and provide the results and checkpoints on KITTI dataset.
 | above @ Pedestrian|||2x|8.12||68.1|59.1|59.5|51.5| |
 | above @ Cyclist|||2x|8.12||60.7|59.5|58.4|57.3| |
 
-Note: See more details about metrics and data split on Waymo [HERE](https://github.com/open-mmlab/mmdetection3d/tree/master/configs/pointpillars). For implementation details, we basically follow the original settings. All of these results are achieved without bells-and-whistles, e.g. ensemble, multi-scale training and test augmentation.
+Note:
+
+- See more details about metrics and data split on Waymo [HERE](https://github.com/open-mmlab/mmdetection3d/tree/master/configs/pointpillars). For implementation details, we basically follow the original settings. All of these results are achieved without bells-and-whistles, e.g. ensemble, multi-scale training and test augmentation.
+- `FP16` means Mixed Precision (FP16) is adopted in training.

configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py → configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py

-_base_ = '../second/hv_second_secfpn_6x8_80e_kitti-3d-3class.py'
+_base_ = './hv_second_secfpn_6x8_80e_kitti-3d-3class.py'
 # fp16 settings
 fp16 = dict(loss_scale=512.)

configs/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py → configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py

-_base_ = '../second/hv_second_secfpn_6x8_80e_kitti-3d-car.py'
+_base_ = './hv_second_secfpn_6x8_80e_kitti-3d-car.py'
 # fp16 settings
 fp16 = dict(loss_scale=512.)

configs/second/metafile.yml

@@ -57,3 +57,41 @@ Models:
           mAPH@L1: 61.7
           mAP@L2: 58.9
           mAPH@L2: 55.7
+
+  - Name: hv_second_secfpn_fp16_6x8_80e_kitti-3d-car
+    In Collection: SECOND
+    Config: configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car.py
+    Metadata:
+      Training Techniques:
+        - AdamW
+        - Mixed Precision Training
+      Training Resources: 8x TITAN Xp
+      Training Data: KITTI
+      Training Memory (GB): 2.9
+    Results:
+      - Task: 3D Object Detection
+        Dataset: KITTI
+        Metrics:
+          mAP: 78.72
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car/hv_second_secfpn_fp16_6x8_80e_kitti-3d-car_20200924_211301-1f5ad833.pth
+    Code:
+      Version: v0.7.0
+
+  - Name: hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class
+    In Collection: SECOND
+    Config: configs/second/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class.py
+    Metadata:
+      Training Techniques:
+        - AdamW
+        - Mixed Precision Training
+      Training Resources: 8x TITAN Xp
+      Training Data: KITTI
+      Training Memory (GB): 2.9
+    Results:
+      - Task: 3D Object Detection
+        Dataset: KITTI
+        Metrics:
+          mAP: 67.4
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fp16/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class/hv_second_secfpn_fp16_6x8_80e_kitti-3d-3class_20200925_110059-05f67bdf.pth
+    Code:
+      Version: v0.7.0

docs/model_zoo.md

@@ -89,3 +89,7 @@ Please refer to [SMOKE](https://github.com/open-mmlab/mmdetection3d/tree/v1.0.0.
 ### PGD
 
 Please refer to [PGD](https://github.com/open-mmlab/mmdetection3d/tree/v1.0.0.dev0/configs/pgd) for details. We provide PGD baselines on KITTI and nuScenes dataset.
+
+### Mixed Precision (FP16) Training
+
+Please refer [Mixed Precision (FP16) Training] on PointPillars (https://github.com/open-mmlab/mmdetection3d/tree/v1.0.0.dev0/configs/pointpillars/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py) for details.

docs_zh-CN/model_zoo.md

@@ -91,3 +91,7 @@
 ### PGD
 
 请参考 [PGD](https://github.com/open-mmlab/mmdetection3d/tree/v1.0.0.dev0/configs/pgd) 获取更多细节，我们在 KITTI 和 nuScenes 数据集上给出了相应的结果.
+
+### Mixed Precision (FP16) Training
+
+细节请参考 [Mixed Precision (FP16) Training] 在 PointPillars 训练的样例 (https://github.com/open-mmlab/mmdetection3d/tree/v1.0.0.dev0/configs/pointpillars/hv_pointpillars_fpn_sbn-all_fp16_2x8_2x_nus-3d.py).

model-index.yml

@@ -2,7 +2,6 @@ Import:
   - configs/3dssd/metafile.yml
   - configs/centerpoint/metafile.yml
   - configs/dynamic_voxelization/metafile.yml
-  - configs/fp16/metafile.yml
   - configs/free_anchor/metafile.yml
   - configs/h3dnet/metafile.yml
   - configs/imvotenet/metafile.yml