Only download CLIP on rank 0 when doing eval (#135)

diffusion/evaluation/clean_fid_eval.py

@@ -5,7 +5,7 @@
 
 import json
 import os
-from typing import List, Optional
+from typing import Dict, List, Optional
 
 import clip
 import torch
@@ -50,6 +50,7 @@ class CleanFIDEvaluator:
         precision (str): The precision to use for evaluation. Default: ``'amp_fp16'``.
         prompts (List[str], optional): The prompts to use for image visualtization.
             Default: ``["A shiba inu wearing a blue sweater]``.
+        additional_generate_kwargs (Dict, optional): Additional keyword arguments to pass to the model.generate method.
 
     """
 
@@ -69,7 +70,8 @@ def __init__(self,
                  output_dir: str = '/tmp/',
                  num_samples: Optional[int] = None,
                  precision: str = 'amp_fp16',
-                 prompts: Optional[List[str]] = None):
+                 prompts: Optional[List[str]] = None,
+                 additional_generate_kwargs: Optional[Dict] = None):
         self.model = model
         self.tokenizer: PreTrainedTokenizerBase = model.tokenizer
         self.eval_dataloader = eval_dataloader
@@ -86,6 +88,7 @@ def __init__(self,
         self.num_samples = num_samples if num_samples is not None else float('inf')
         self.precision = precision
         self.prompts = prompts if prompts is not None else ['A shiba inu wearing a blue sweater']
+        self.additional_generate_kwargs = additional_generate_kwargs if additional_generate_kwargs is not None else {}
         self.sdxl = model.sdxl
 
         # Init loggers
@@ -107,7 +110,13 @@ def __init__(self,
         self.clip_metric = self.clip_metric.to(self.device)
 
         # Predownload the CLIP model for computing clip-fid
-        _, _ = clip.load('ViT-B/32', device=self.device)
+        clip_url = 'https://openaipublic.azureedge.net/clip/models/40d365715913c9da98579312b702a82c18be219cc2a73407c4526f58eba950af/ViT-B-32.pt'
+        clip_name = os.path.basename(clip_url)
+        clip_path = os.path.expanduser('~/.cache/clip')
+        if dist.get_local_rank() == 0:
+            clip.clip._download(clip_url, clip_path)
+        with dist.local_rank_zero_download_and_wait(os.path.join(clip_path, clip_name)):
+            clip.load('ViT-B/32', device=self.device)
 
     def _generate_images(self, guidance_scale: float):
         """Core image generation function. Generates images at a given guidance scale.
@@ -156,7 +165,8 @@ def _generate_images(self, guidance_scale: float):
                                                        seed=seed,
                                                        crop_params=crop_params,
                                                        input_size_params=input_size_params,
-                                                       progress_bar=False)  # type: ignore
+                                                       progress_bar=False,
+                                                       **self.additional_generate_kwargs)  # type: ignore
             # Get the prompts from the tokens
             text_captions = self.tokenizer.batch_decode(captions, skip_special_tokens=True)
             self.clip_metric.update((generated_images * 255).to(torch.uint8), text_captions)
@@ -233,7 +243,8 @@ def _generate_images_from_prompts(self, guidance_scale: float):
                                                        height=self.size,
                                                        width=self.size,
                                                        guidance_scale=guidance_scale,
-                                                       seed=self.seed)  # type: ignore
+                                                       seed=self.seed,
+                                                       **self.additional_generate_kwargs)  # type: ignore
         else:
             generated_images = []
         return generated_images