MaskRCNN use smaller LR and add 128GPU config

PyTorch/Segmentation/MaskRCNN/pytorch/configs/e2e_mask_rcnn_R_50_FPN_1x_128GPU_4bs.yaml

@@ -0,0 +1,47 @@
+MODEL:
+  META_ARCHITECTURE: "GeneralizedRCNN"
+  WEIGHT: "catalog://ImageNetPretrained/MSRA/R-50"
+  BACKBONE:
+    CONV_BODY: "R-50-FPN"
+    OUT_CHANNELS: 256
+  RPN:
+    USE_FPN: True
+    ANCHOR_STRIDE: (4, 8, 16, 32, 64)
+    PRE_NMS_TOP_N_TRAIN: 2000
+    PRE_NMS_TOP_N_TEST: 1000
+    POST_NMS_TOP_N_TEST: 1000
+    FPN_POST_NMS_TOP_N_TEST: 1000
+  ROI_HEADS:
+    USE_FPN: True
+  ROI_BOX_HEAD:
+    POOLER_RESOLUTION: 7
+    POOLER_SCALES: (0.25, 0.125, 0.0625, 0.03125)
+    POOLER_SAMPLING_RATIO: 2
+    FEATURE_EXTRACTOR: "FPN2MLPFeatureExtractor"
+    PREDICTOR: "FPNPredictor"
+  ROI_MASK_HEAD:
+    POOLER_SCALES: (0.25, 0.125, 0.0625, 0.03125)
+    FEATURE_EXTRACTOR: "MaskRCNNFPNFeatureExtractor"
+    PREDICTOR: "MaskRCNNC4Predictor"
+    POOLER_RESOLUTION: 14
+    POOLER_SAMPLING_RATIO: 2
+    RESOLUTION: 28
+    SHARE_BOX_FEATURE_EXTRACTOR: False
+  MASK_ON: True
+DATASETS:
+  TRAIN: ("coco_2017_train",)
+  TEST: ("coco_2017_val",)
+DATALOADER:
+  SIZE_DIVISIBILITY: 32
+SOLVER:
+  BASE_LR: 0.0005
+  WEIGHT_DECAY: 0.0001
+  STEPS: (4320, 5760)
+  MAX_ITER: 2000
+  IMS_PER_BATCH: 256
+  WARMUP_FACTOR: 0.001
+  WARMUP_ITERS: 1000
+TEST:
+  IMS_PER_BATCH: 64
+
+

PyTorch/Segmentation/MaskRCNN/pytorch/configs/e2e_mask_rcnn_R_50_FPN_1x_32GPU_4bs.yaml

@@ -34,10 +34,10 @@ DATASETS:
 DATALOADER:
   SIZE_DIVISIBILITY: 32
 SOLVER:
-  BASE_LR: 0.02
+  BASE_LR: 0.002
   WEIGHT_DECAY: 0.0001
   STEPS: (4320, 5760)
-  MAX_ITER: 1000
+  MAX_ITER: 2000
   IMS_PER_BATCH: 128
   WARMUP_FACTOR: 0.001
   WARMUP_ITERS: 1000

PyTorch/Segmentation/MaskRCNN/pytorch/configs/e2e_mask_rcnn_R_50_FPN_1x_64GPU_4bs.yaml

@@ -34,7 +34,7 @@ DATASETS:
 DATALOADER:
   SIZE_DIVISIBILITY: 32
 SOLVER:
-  BASE_LR: 0.04
+  BASE_LR: 0.001
   WEIGHT_DECAY: 0.0001
   STEPS: (4320, 5760)
   MAX_ITER: 2000