Updates to be able to train flownet2 in ExtensibleTrainer

Only supports basic losses for now, though.

codes/data/paired_frame_dataset.py

@@ -35,7 +35,7 @@ class PairedFrameDataset(BaseUnsupervisedImageDataset):
         # Convert to torch tensor
         hq = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(hs), (0, 3, 1, 2)))).float()
         hq_ref = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(hrs), (0, 3, 1, 2)))).float()
-        hq_mask = torch.from_numpy(np.ascontiguousarray(np.stack(hms))).unsqueeze(dim=1)
+        hq_mask = torch.from_numpy(np.ascontiguousarray(np.stack(hms))).squeeze().unsqueeze(dim=1)
         hq_ref = torch.cat([hq_ref, hq_mask], dim=1)
         lq = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(ls), (0, 3, 1, 2)))).float()
         lq_ref = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(lrs), (0, 3, 1, 2)))).float()

codes/models/base_model.py

@@ -101,6 +101,11 @@ class BaseModel():
         if isinstance(network, nn.DataParallel) or isinstance(network, DistributedDataParallel):
             network = network.module
         load_net = torch.load(load_path)
+
+        # Support loading torch.save()s for whole models as well as just state_dicts.
+        if 'state_dict' in load_net:
+            load_net = load_net['state_dict']
+
         load_net_clean = OrderedDict()  # remove unnecessary 'module.'
         for k, v in load_net.items():
             if k.startswith('module.'):

codes/models/networks.py

@@ -87,10 +87,12 @@ def define_G(opt, net_key='network_G', scale=None):
         netG = chained.MultifacetedChainedEmbeddingGen(depth=opt_net['depth'], scale=scale)
     elif which_model == "flownet2":
         from models.flownet2.models import FlowNet2
-        ld = torch.load(opt_net['load_path'])
+        ld = 'load_path' in opt_net.keys()
-        args = munch.Munch({'fp16': False, 'rgb_max': 1.0})
+        args = munch.Munch({'fp16': False, 'rgb_max': 1.0, 'checkpoint': not ld})
         netG = FlowNet2(args)
-        netG.load_state_dict(ld['state_dict'])
+        if ld:
+            sd = torch.load(opt_net['load_path'])
+            netG.load_state_dict(sd['state_dict'])
     elif which_model == "backbone_encoder":
         netG = SwitchedGen_arch.BackboneEncoder(pretrained_backbone=opt_net['pretrained_spinenet'])
     elif which_model == "backbone_encoder_no_ref":

codes/models/steps/injectors.py

@@ -14,6 +14,9 @@ def create_injector(opt_inject, env):
     elif 'progressive_' in type:
         from models.steps.progressive_zoom import create_progressive_zoom_injector
         return create_progressive_zoom_injector(opt_inject, env)
+    elif 'stereoscopic_' in type:
+        from models.steps.stereoscopic import create_stereoscopic_injector
+        return create_stereoscopic_injector(opt_inject, env)
     elif type == 'generator':
         return ImageGeneratorInjector(opt_inject, env)
     elif type == 'discriminator':
@@ -42,6 +45,8 @@ def create_injector(opt_inject, env):
         return ConstantInjector(opt_inject, env)
     elif type == 'fft':
         return ImageFftInjector(opt_inject, env)
+    elif type == 'extract_indices':
+        return IndicesExtractor(opt_inject, env)
     else:
         raise NotImplementedError
 
@@ -299,3 +304,17 @@ class ImageFftInjector(Injector):
             im = torch.irfft(fftim, signal_ndim=2, normalized=True)
             return {self.output: im}
 
+
+class IndicesExtractor(Injector):
+    def __init__(self, opt, env):
+        super(IndicesExtractor, self).__init__(opt, env)
+        self.dim = opt['dim']
+        assert self.dim == 1  # Honestly not sure how to support an abstract dim here, so just add yours when needed.
+
+    def forward(self, state):
+        results = {}
+        for i, o in enumerate(self.output):
+            if self.dim == 1:
+                results[o] = state[self.input][:, i]
+        return results
+

codes/models/steps/stereoscopic.py

@@ -0,0 +1,22 @@
+import torch
+from torch.cuda.amp import autocast
+from models.flownet2.networks.resample2d_package.resample2d import Resample2d
+from models.steps.injectors import Injector
+
+
+def create_stereoscopic_injector(opt, env):
+    type = opt['type']
+    if type == 'stereoscopic_resample':
+        return ResampleInjector(opt, env)
+    return None
+
+
+class ResampleInjector(Injector):
+    def __init__(self, opt, env):
+        super(ResampleInjector, self).__init__(opt, env)
+        self.resample = Resample2d()
+        self.flow = opt['flowfield']
+
+    def forward(self, state):
+        with autocast(enabled=False):
+            return {self.output: self.resample(state[self.input], state[self.flow])}

codes/train.py

@@ -278,7 +278,7 @@ class Trainer:
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
-    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_prog_imgset_multifaceted_chained.yml')
+    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_3dflow_vr_flownet.yml')
     parser.add_argument('--launcher', choices=['none', 'pytorch'], default='none', help='job launcher')
     args = parser.parse_args()
     opt = option.parse(args.opt, is_train=True)

codes/utils/util.py

@@ -51,6 +51,14 @@ def checkpoint(fn, *args):
     else:
         return fn(*args)
 
+# A fancy alternative to if <flag> checkpoint() else <call>
+def possible_checkpoint(enabled, fn, *args):
+    opt_en = loaded_options['checkpointing_enabled'] if 'checkpointing_enabled' in loaded_options.keys() else True
+    if enabled and opt_en:
+        return torch.utils.checkpoint.checkpoint(fn, *args)
+    else:
+        return fn(*args)
+
 def get_timestamp():
     return datetime.now().strftime('%y%m%d-%H%M%S')