Add FID evaluator for diffusion models

codes/test.py

@@ -58,7 +58,7 @@ if __name__ == "__main__":
     torch.backends.cudnn.benchmark = True
     want_metrics = False
     parser = argparse.ArgumentParser()
-    parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/test_diffusion_unet_sm.yml')
+    parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/test_diffusion_unet.yml')
     opt = option.parse(parser.parse_args().opt, is_train=False)
     opt = option.dict_to_nonedict(opt)
     utils.util.loaded_options = opt

codes/trainer/ExtensibleTrainer.py

@@ -143,6 +143,7 @@ class ExtensibleTrainer(BaseModel):
         # Replace the env networks with the wrapped networks
         self.env['generators'] = self.netsG
         self.env['discriminators'] = self.netsD
+        self.env['emas'] = self.emas
 
         self.print_network()  # print network
         self.load()  # load networks from save states as needed

codes/trainer/eval/evaluator.py

@@ -17,7 +17,7 @@ class Evaluator:
 
 
 def format_evaluator_name(name):
-    # Formats by converting from CamelCase to snake_case and removing trailing "_injector"
+    # Formats by converting from CamelCase to snake_case and removing trailing "_evaluator"
     name = re.sub('(.)([A-Z][a-z]+)', r'\1_\2', name)
     name = re.sub('([a-z0-9])([A-Z])', r'\1_\2', name).lower()
     return name.replace("_evaluator", "")

codes/trainer/eval/sr_diffusion_fid.py

@@ -0,0 +1,50 @@
+import os
+import torch
+import os.path as osp
+import torchvision
+from torch.nn.functional import interpolate
+from tqdm import tqdm
+
+import trainer.eval.evaluator as evaluator
+
+from pytorch_fid import fid_score
+from data import create_dataset
+from torch.utils.data import DataLoader
+
+from trainer.injectors.gaussian_diffusion_injector import GaussianDiffusionInferenceInjector
+from utils.util import opt_get
+
+
+class SrDiffusionFidEvaluator(evaluator.Evaluator):
+    def __init__(self, model, opt_eval, env):
+        super().__init__(model, opt_eval, env)
+        self.batch_sz = opt_eval['batch_size']
+        self.fid_batch_size = opt_get(opt_eval, ['fid_batch_size'], 64)
+        assert self.batch_sz is not None
+        self.dataset = create_dataset(opt_eval['dataset'])
+        self.fid_real_samples = opt_eval['dataset']['paths']  # This is assumed to exist for the given dataset.
+        assert isinstance(self.fid_real_samples, str)
+        self.dataloader = DataLoader(self.dataset, self.batch_sz, shuffle=False, num_workers=1)
+        self.gd = GaussianDiffusionInferenceInjector(opt_eval['diffusion_params'], env)
+        self.out_key = opt_eval['diffusion_params']['out']
+
+    def perform_eval(self):
+        fid_fake_path = osp.join(self.env['base_path'], "..", "fid", str(self.env["step"]))
+        os.makedirs(fid_fake_path, exist_ok=True)
+        counter = 0
+        for batch in tqdm(self.dataloader):
+            batch = {k: v.to(self.env['device']) if isinstance(v, torch.Tensor) else v for k, v in batch.items()}
+            gen = self.gd(batch)[self.out_key]
+
+            # All gather if we're in distributed mode.
+            if torch.distributed.is_available() and torch.distributed.is_initialized():
+                gather_list = [torch.zeros_like(gen) for _ in range(torch.distributed.get_world_size())]
+                torch.distributed.all_gather(gather_list, gen)
+                gen = torch.cat(gather_list, dim=0)
+
+            for b in range(self.batch_sz):
+                torchvision.utils.save_image(gen[b], osp.join(fid_fake_path, "%i_.png" % (counter)))
+                counter += 1
+
+        return {"fid": fid_score.calculate_fid_given_paths([self.fid_real_samples, fid_fake_path], self.fid_batch_size,
+                                                           True, 2048)}

codes/trainer/injectors/gaussian_diffusion_injector.py

@@ -48,9 +48,13 @@ class GaussianDiffusionInferenceInjector(Injector):
                                                                  [opt_get(opt, ['respaced_timestep_spacing'], opt['beta_schedule']['num_diffusion_timesteps'])])
         self.diffusion = SpacedDiffusion(**opt['diffusion_args'])
         self.model_input_keys = opt_get(opt, ['model_input_keys'], [])
+        self.use_ema_model = opt_get(opt, ['use_ema'], False)
 
     def forward(self, state):
-        gen = self.env['generators'][self.opt['generator']]
+        if self.use_ema_model:
+            gen = self.env['emas'][self.opt['generator']]
+        else:
+            gen = self.env['generators'][self.opt['generator']]
         model_inputs = {k: state[v][:self.output_batch_size] for k, v in self.model_input_keys.items()}
         gen.eval()
         with torch.no_grad():