Translating Detectron1 config file to Detectron2?

[cooldecola]

Hey guys, I have a yaml file in detectron that I used to train a model. I am now trying to convert to detectron2 and can't seem to find the same parameters in detectron2.

This is the yaml file in detectron:

MODEL: 
  TYPE: generalized-rcnn
  CONV_BODY: FPN.add_fpn_ResNet50_conv5_body
  NUM_CLASSES: 81
  FASTER_RCNN: True

NUM_GPUS: 1

SOLVER: 
  WEIGHT_DECAY: 0.0001
  LR_POLICY: steps_with_decay
  BASE_LR: 0.0025
  GAMMA: 0.1
  MAX_ITER: 5400
  STEPS: [0, 3000, 4000]

FPN: 
  FPN_ON: True
  MULTILEVEL_ROIS: True
  MULTILEVEL_RPN: True

FAST_RCNN: 
  ROI_BOX_HEAD: fast_rcnn_heads.add_roi_2mlp_head
  ROI_XFORM_METHOD: RoIAlign
  ROI_XFORM_RESOLUTION: 7
  ROI_XFORM_SAMPLING_RATIO: 2

TRAIN: 
  WEIGHTS: https://dl.fbaipublicfiles.com/detectron/ImageNetPretrained/MSRA/R-50.pkl
  DATASETS: ('d1', )
  SCALES: (500,)
  MAX_SIZE: 833
  BATCH_SIZE_PER_IM: 256
  RPN_PRE_NMS_TOP_N: 2000

D1 = detectron1
D2 = detectron2

Problem 1

In D1 I set CONV_BODY: FPN.add_fpn_ResNet50_conv5_body however in D2 I have it set as cfg.merge_from_file(model_zoo.get_config_file("Base-RCNN-FPN.yaml")). Inside Base-RCNN-FPN.yaml I see

BACKBONE:
    NAME: "build_resnet_fpn_backbone"

Are these backbones the same? build_resnet_fpn_backbone does not specify if it is ResNet50

Problem 2

How do I set how many GPU's ? In D1 I have NUM_GPUS: 1 but nothing similar in D2?

Problem 3

Under the FPN section of my yaml config file I have set

FPN: 
  FPN_ON: True
  MULTILEVEL_ROIS: True
  MULTILEVEL_RPN: True

I don't have the option to use FPN for RoI transforms or using FPN for RPN. Is there a way to do this?

Problem 4

Same issue with the FAST_RCNN options. I cannot find similar options that translate over.

The following is what I have implemented so far in D2 but I'm not sure if this is correct:

cfg.MODEL.ROI_BOX_HEAD.NAME = "FastRCNNConvFCHead"
cfg.MODEL.ROI_BOX_HEAD.BBOX_REG_LOSS_TYPE = "smooth_li"
cfg.MODEL.ROI_BOX_HEAD.POOLER_RESOLUTION = 7
cfg.MODEL.ROI_BOX_HEAD.POOLER_SAMPLER_RATIO = 2

Last Problem

I have set BATCH_SIZE_PER_IM = 256 in D1.
Is this the same as cfg.MODEL.BATCH_SIZE_PER_IMAGE = 256 ?
The reason I am asking is that in D1 the BATCH_SIZE_PER_IM falls under the training category, however in D2 it's under the model category. I think they should be the same but I just wanted to clarify.

Any help would be greatly appreciated, thanks in advance! :)