Cleanup EfficientNet/MobileNetV3 feature extraction a bit, only two tap locations now, small mobilenetv3 models work

timm/models/efficientnet.py

@@ -416,12 +416,6 @@ class EfficientNetFeatures(nn.Module):
                  se_kwargs=None, norm_layer=nn.BatchNorm2d, norm_kwargs=None):
         super(EfficientNetFeatures, self).__init__()
         norm_kwargs = norm_kwargs or {}
-
-        # TODO only create stages needed, currently all stages are created regardless of out_indices
-        num_stages = max(out_indices) + 1
-
-        self.out_indices = out_indices
-        self.feature_location = feature_location
         self.drop_rate = drop_rate
         self._in_chs = in_chans
 
@@ -439,14 +433,10 @@ class EfficientNetFeatures(nn.Module):
             norm_layer, norm_kwargs, drop_path_rate, feature_location=feature_location, verbose=_DEBUG)
         self.blocks = nn.Sequential(*builder(self._in_chs, block_args))
         self.feature_info = FeatureInfo(builder.features, out_indices)
-        self._stage_to_feature_idx = {
+        self._stage_out_idx = {v['stage']: i for i, v in enumerate(self.feature_info) if i in out_indices}
-            v['stage_idx']: fi for fi, v in enumerate(self.feature_info) if fi in self.out_indices}
         self._in_chs = builder.in_chs
 
         efficientnet_init_weights(self)
-        if _DEBUG:
-            for fi, v in enumerate(self.feature_info):
-                print('Feature idx: {}: Name: {}, Channels: {}'.format(fi, v['module'], v['num_chs']))
 
         # Register feature extraction hooks with FeatureHooks helper
         self.feature_hooks = None
@@ -460,14 +450,17 @@ class EfficientNetFeatures(nn.Module):
         x = self.act1(x)
         if self.feature_hooks is None:
             features = []
+            if 0 in self._stage_out_idx:
+                features.append(x)  # add stem out
             for i, b in enumerate(self.blocks):
                 x = b(x)
-                if i in self._stage_to_feature_idx:
+                if i + 1 in self._stage_out_idx:
                     features.append(x)
             return features
         else:
             self.blocks(x)
-            return self.feature_hooks.get_output(x.device)
+            out = self.feature_hooks.get_output(x.device)
+            return list(out.values())
 
 
 def _create_effnet(model_kwargs, variant, pretrained=False):

timm/models/efficientnet_blocks.py

@@ -128,10 +128,9 @@ class ConvBnAct(nn.Module):
         self.act1 = act_layer(inplace=True)
 
     def feature_info(self, location):
-        if location == 'expansion' or location == 'depthwise':
+        if location == 'expansion':  # output of conv after act, same as block coutput
-            # no expansion or depthwise this block, use act after conv
             info = dict(module='act1', hook_type='forward', num_chs=self.conv.out_channels)
-        else:  # location == 'bottleneck'
+        else:  # location == 'bottleneck', block output
             info = dict(module='', hook_type='', num_chs=self.conv.out_channels)
         return info
 
@@ -175,12 +174,9 @@ class DepthwiseSeparableConv(nn.Module):
         self.act2 = act_layer(inplace=True) if self.has_pw_act else nn.Identity()
 
     def feature_info(self, location):
-        if location == 'expansion':
+        if location == 'expansion':  # after SE, input to PW
-            # no expansion in this block, use depthwise, before SE
-            info = dict(module='act1', hook_type='forward', num_chs=self.conv_pw.in_channels)
-        elif location == 'depthwise':  # after SE
             info = dict(module='conv_pw', hook_type='forward_pre', num_chs=self.conv_pw.in_channels)
-        else:  # location == 'bottleneck'
+        else:  # location == 'bottleneck', block output
             info = dict(module='', hook_type='', num_chs=self.conv_pw.out_channels)
         return info
 
@@ -245,11 +241,9 @@ class InvertedResidual(nn.Module):
         self.bn3 = norm_layer(out_chs, **norm_kwargs)
 
     def feature_info(self, location):
-        if location == 'expansion':
+        if location == 'expansion':  # after SE, input to PWL
-            info = dict(module='act1', hook_type='forward', num_chs=self.conv_pw.in_channels)
-        elif location == 'depthwise':  # after SE
             info = dict(module='conv_pwl', hook_type='forward_pre', num_chs=self.conv_pwl.in_channels)
-        else:  # location == 'bottleneck'
+        else:  # location == 'bottleneck', block output
             info = dict(module='', hook_type='', num_chs=self.conv_pwl.out_channels)
         return info
 
@@ -370,12 +364,9 @@ class EdgeResidual(nn.Module):
         self.bn2 = norm_layer(out_chs, **norm_kwargs)
 
     def feature_info(self, location):
-        if location == 'expansion':
+        if location == 'expansion':  # after SE, before PWL
-            info = dict(module='act1', hook_type='forward', num_chs=self.conv_exp.out_channels)
-        elif location == 'depthwise':
-            # there is no depthwise, take after SE, before PWL
             info = dict(module='conv_pwl', hook_type='forward_pre', num_chs=self.conv_pwl.in_channels)
-        else:  # location == 'bottleneck'
+        else:  # location == 'bottleneck', block output
             info = dict(module='', hook_type='', num_chs=self.conv_pwl.out_channels)
         return info
 

timm/models/efficientnet_builder.py

@@ -1,7 +1,6 @@
 import logging
 import math
 import re
-from collections import OrderedDict
 from copy import deepcopy
 
 import torch.nn as nn
@@ -12,6 +11,11 @@ from .layers import CondConv2d, get_condconv_initializer
 __all__ = ["EfficientNetBuilder", "decode_arch_def", "efficientnet_init_weights"]
 
 
+def _log_info_if(msg, condition):
+    if condition:
+        logging.info(msg)
+
+
 def _parse_ksize(ss):
     if ss.isdigit():
         return int(ss)
@@ -219,8 +223,12 @@ class EfficientNetBuilder:
         self.norm_layer = norm_layer
         self.norm_kwargs = norm_kwargs
         self.drop_path_rate = drop_path_rate
+        if feature_location == 'depthwise':
+            # old 'depthwise' mode renamed 'expansion' to match TF impl, old expansion mode didn't make sense
+            logging.warning("feature_location=='depthwise' is deprecated, using 'expansion'")
+            feature_location = 'expansion'
         self.feature_location = feature_location
-        assert feature_location in ('bottleneck', 'depthwise', 'expansion', '')
+        assert feature_location in ('bottleneck', 'expansion', '')
         self.verbose = verbose
 
         # state updated during build, consumed by model
@@ -247,8 +255,7 @@ class EfficientNetBuilder:
         if bt == 'ir':
             ba['drop_path_rate'] = drop_path_rate
             ba['se_kwargs'] = self.se_kwargs
-            if self.verbose:
+            _log_info_if('  InvertedResidual {}, Args: {}'.format(block_idx, str(ba)), self.verbose)
-                logging.info('  InvertedResidual {}, Args: {}'.format(block_idx, str(ba)))
             if ba.get('num_experts', 0) > 0:
                 block = CondConvResidual(**ba)
             else:
@@ -256,18 +263,15 @@ class EfficientNetBuilder:
         elif bt == 'ds' or bt == 'dsa':
             ba['drop_path_rate'] = drop_path_rate
             ba['se_kwargs'] = self.se_kwargs
-            if self.verbose:
+            _log_info_if('  DepthwiseSeparable {}, Args: {}'.format(block_idx, str(ba)), self.verbose)
-                logging.info('  DepthwiseSeparable {}, Args: {}'.format(block_idx, str(ba)))
             block = DepthwiseSeparableConv(**ba)
         elif bt == 'er':
             ba['drop_path_rate'] = drop_path_rate
             ba['se_kwargs'] = self.se_kwargs
-            if self.verbose:
+            _log_info_if('  EdgeResidual {}, Args: {}'.format(block_idx, str(ba)), self.verbose)
-                logging.info('  EdgeResidual {}, Args: {}'.format(block_idx, str(ba)))
             block = EdgeResidual(**ba)
         elif bt == 'cn':
-            if self.verbose:
+            _log_info_if('  ConvBnAct {}, Args: {}'.format(block_idx, str(ba)), self.verbose)
-                logging.info('  ConvBnAct {}, Args: {}'.format(block_idx, str(ba)))
             block = ConvBnAct(**ba)
         else:
             assert False, 'Uknkown block type (%s) while building model.' % bt
@@ -279,64 +283,55 @@ class EfficientNetBuilder:
         """ Build the blocks
         Args:
             in_chs: Number of input-channels passed to first block
-            model_block_args: A list of lists, outer list defines stages, inner
+            model_block_args: A list of lists, outer list defines stacks (block stages), inner
                 list contains strings defining block configuration(s)
         Return:
              List of block stacks (each stack wrapped in nn.Sequential)
         """
-        if self.verbose:
+        _log_info_if('Building model trunk with %d stages...' % len(model_block_args), self.verbose)
-            logging.info('Building model trunk with %d stages...' % len(model_block_args))
         self.in_chs = in_chs
         total_block_count = sum([len(x) for x in model_block_args])
         total_block_idx = 0
         current_stride = 2
         current_dilation = 1
         stages = []
-        # outer list of block_args defines the stacks ('stages' by some conventions)
+        if model_block_args[0][0]['stride'] > 1:
-        for stage_idx, stage_block_args in enumerate(model_block_args):
+            # if the first block starts with a stride, we need to extract first level feat from stem
-            last_stack = stage_idx == (len(model_block_args) - 1)
+            feature_info = dict(
-            if self.verbose:
+                module='act1', num_chs=in_chs, stage=0, reduction=current_stride,
-                logging.info('Stack: {}'.format(stage_idx))
+                hook_type='forward' if self.feature_location != 'bottleneck' else '')
-            assert isinstance(stage_block_args, list)
+            self.features.append(feature_info)
+
+        # outer list of block_args defines the stacks
+        for stack_idx, stack_args in enumerate(model_block_args):
+            last_stack = stack_idx + 1 == len(model_block_args)
+            _log_info_if('Stack: {}'.format(stack_idx), self.verbose)
+            assert isinstance(stack_args, list)
 
             blocks = []
-            # each stack (stage) contains a list of block arguments
+            # each stack (stage of blocks) contains a list of block arguments
-            for block_idx, block_args in enumerate(stage_block_args):
+            for block_idx, block_args in enumerate(stack_args):
-                last_block = block_idx == (len(stage_block_args) - 1)
+                last_block = block_idx + 1 == len(stack_args)
-                extract_features = ''  # No features extracted
+                _log_info_if(' Block: {}'.format(block_idx), self.verbose)
-                if self.verbose:
+
-                    logging.info(' Block: {}'.format(block_idx))
-
-                # Sort out stride, dilation, and feature extraction details
                 assert block_args['stride'] in (1, 2)
-                if block_idx >= 1:
+                if block_idx >= 1:   # only the first block in any stack can have a stride > 1
-                    # only the first block in any stack can have a stride > 1
                     block_args['stride'] = 1
 
-                do_extract = False
+                extract_features = False
-                if self.feature_location == 'bottleneck' or self.feature_location == 'depthwise':
                 if last_block:
-                        next_stage_idx = stage_idx + 1
+                    next_stack_idx = stack_idx + 1
-                        if next_stage_idx >= len(model_block_args):
+                    extract_features = next_stack_idx >= len(model_block_args) or \
-                            do_extract = True
+                        model_block_args[next_stack_idx][0]['stride'] > 1
-                        else:
-                            do_extract = model_block_args[next_stage_idx][0]['stride'] > 1
-                elif self.feature_location == 'expansion':
-                    if block_args['stride'] > 1 or (last_stack and last_block):
-                        do_extract = True
-                if do_extract:
-                    extract_features = self.feature_location
 
                 next_dilation = current_dilation
-                next_output_stride = current_stride
                 if block_args['stride'] > 1:
                     next_output_stride = current_stride * block_args['stride']
                     if next_output_stride > self.output_stride:
                         next_dilation = current_dilation * block_args['stride']
                         block_args['stride'] = 1
-                        if self.verbose:
+                        _log_info_if('  Converting stride to dilation to maintain output_stride=={}'.format(
-                            logging.info('  Converting stride to dilation to maintain output_stride=={}'.format(
+                            self.output_stride), self.verbose)
-                                self.output_stride))
                     else:
                         current_stride = next_output_stride
                 block_args['dilation'] = current_dilation
@@ -349,15 +344,11 @@ class EfficientNetBuilder:
 
                 # stash feature module name and channel info for model feature extraction
                 if extract_features:
-                    feature_info = block.feature_info(extract_features)
+                    feature_info = dict(
-                    module_name = f'blocks.{stage_idx}.{block_idx}'
+                        stage=stack_idx + 1, reduction=current_stride, **block.feature_info(self.feature_location))
-                    if 'module' in feature_info and feature_info['module']:
+                    module_name = f'blocks.{stack_idx}.{block_idx}'
-                        feature_info['module'] = '.'.join([module_name, feature_info['module']])
+                    leaf_name = feature_info.get('module', '')
-                    else:
+                    feature_info['module'] = '.'.join([module_name, leaf_name]) if leaf_name else module_name
-                        feature_info['module'] = module_name
-                    feature_info['stage_idx'] = stage_idx
-                    feature_info['block_idx'] = block_idx
-                    feature_info['reduction'] = current_stride
                     self.features.append(feature_info)
 
                 total_block_idx += 1  # incr global block idx (across all stacks)

timm/models/mobilenetv3.py

@@ -162,11 +162,6 @@ class MobileNetV3Features(nn.Module):
                  norm_layer=nn.BatchNorm2d, norm_kwargs=None):
         super(MobileNetV3Features, self).__init__()
         norm_kwargs = norm_kwargs or {}
-
-        # TODO only create stages needed, currently all stages are created regardless of out_indices
-        num_stages = max(out_indices) + 1
-
-        self.out_indices = out_indices
         self.drop_rate = drop_rate
         self._in_chs = in_chans
 
@@ -183,14 +178,10 @@ class MobileNetV3Features(nn.Module):
             norm_layer, norm_kwargs, drop_path_rate, feature_location=feature_location, verbose=_DEBUG)
         self.blocks = nn.Sequential(*builder(self._in_chs, block_args))
         self.feature_info = FeatureInfo(builder.features, out_indices)
-        self._stage_to_feature_idx = {
+        self._stage_out_idx = {v['stage']: i for i, v in enumerate(self.feature_info) if i in out_indices}
-            v['stage_idx']: fi for fi, v in enumerate(self.feature_info) if fi in self.out_indices}
         self._in_chs = builder.in_chs
 
         efficientnet_init_weights(self)
-        if _DEBUG:
-            for fi, v in enumerate(self.feature_info):
-                print('Feature idx: {}: Name: {}, Channels: {}'.format(fi, v['module'], v['num_chs']))
 
         # Register feature extraction hooks with FeatureHooks helper
         self.feature_hooks = None
@@ -204,9 +195,11 @@ class MobileNetV3Features(nn.Module):
         x = self.act1(x)
         if self.feature_hooks is None:
             features = []
+            if 0 in self._stage_out_idx:
+                features.append(x)  # add stem out
             for i, b in enumerate(self.blocks):
                 x = b(x)
-                if i in self._stage_to_feature_idx:
+                if i + 1 in self._stage_out_idx:
                     features.append(x)
             return features
         else: