Model weights might be out of date?

[hansihe]

I am on 6791682.

Running the following code:

runner = GeneralizedRCNNRunner()
cfg = runner.get_default_cfg()

cfg.merge_from_file("/content/faster_rcnn_fbnetv3a_C4.yaml") # This is the config file from the same revision, just working around an issue with colab.
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("faster_rcnn_fbnetv3a_C4.yaml")

predictor = DefaultPredictor(cfg)

Gives the following output:

INFO:d2go.modeling.backbone.fbnet_v2:Build FBNet using unified arch_def:
trunk
- {'block_op': 'conv_k3', 'block_cfg': {'out_channels': 16, 'stride': 2}, 'stage_idx': 0, 'block_idx': 0}
- {'block_op': 'ir_k3', 'block_cfg': {'out_channels': 16, 'stride': 1, 'expansion': 1, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 1}
- {'block_op': 'ir_k3', 'block_cfg': {'out_channels': 16, 'stride': 1, 'expansion': 1, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 2}
- {'block_op': 'ir_k5', 'block_cfg': {'out_channels': 24, 'stride': 2, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 1, 'block_idx': 0}
- {'block_op': 'ir_k5', 'block_cfg': {'out_channels': 24, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 1, 'block_idx': 1}
- {'block_op': 'ir_k5', 'block_cfg': {'out_channels': 24, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 1, 'block_idx': 2}
- {'block_op': 'ir_k5', 'block_cfg': {'out_channels': 24, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 1, 'block_idx': 3}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 32, 'stride': 2, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 2, 'block_idx': 0}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 32, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 2, 'block_idx': 1}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 32, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 2, 'block_idx': 2}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 32, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 2, 'block_idx': 3}
- {'block_op': 'ir_k5', 'block_cfg': {'out_channels': 64, 'stride': 2, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 0}
- {'block_op': 'ir_k3', 'block_cfg': {'out_channels': 64, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 1}
- {'block_op': 'ir_k3', 'block_cfg': {'out_channels': 64, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 2}
- {'block_op': 'ir_k3', 'block_cfg': {'out_channels': 64, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 3}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 4}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 5}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 6}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 7}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 8}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 3, 'block_idx': 9}
WARNING:mobile_cv.arch.utils.helper:Arguments ['width_divisor', 'dw_skip_bnrelu', 'zero_last_bn_gamma'] skipped for op Conv2d
INFO:d2go.modeling.backbone.fbnet_v2:Build FBNet using unified arch_def:
rpn
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 0}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 1}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 2}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 3}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 112, 'stride': 1, 'expansion': 3, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 4}
INFO:d2go.modeling.backbone.fbnet_v2:Build FBNet using unified arch_def:
bbox
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 184, 'stride': 2, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 0}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 184, 'stride': 1, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 1}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 184, 'stride': 1, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 2}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 184, 'stride': 1, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 3}
- {'block_op': 'ir_k3_se', 'block_cfg': {'out_channels': 184, 'stride': 1, 'expansion': 4, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 4}
- {'block_op': 'ir_k5_se', 'block_cfg': {'out_channels': 200, 'stride': 1, 'expansion': 6, 'less_se_channels': False}, 'stage_idx': 0, 'block_idx': 5}
INFO:fvcore.common.checkpoint:[Checkpointer] Loading from https://mobile-cv.s3-us-west-2.amazonaws.com/d2go/models/268421013/model_final.pth ...
WARNING:fvcore.common.checkpoint:The checkpoint state_dict contains keys that are not used by the model:
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_0.pw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_0.pw.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_0.dw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_0.pwl.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_0.pwl.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_1.pw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_1.pw.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_1.dw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_1.pwl.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_1.pwl.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_2.pw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_2.pw.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_2.dw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_2.pwl.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_2.pwl.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_3.pw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_3.pw.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_3.dw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_3.pwl.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_3.pwl.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_4.pw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_4.pw.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_4.dw.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_4.pwl.conv.weight
  roi_heads.mask_head.feature_extractor.0.fbnetv2_0_4.pwl.bn.{bias, num_batches_tracked, running_mean, running_var, weight}
  roi_heads.mask_head.predictor.mask_fcn_logits.{bias, weight}

As I understand it the last couple of times an error like this has occurred it has been an issue with out of date weights, could this be the case here as well?

[wat3rBro]

It's loading a maskrcnn weight file into a faster-rcnn model, as you can see the un-used parameters are all starting with "roi_heads".