Fix conversion script

debug_conversion.py

@@ -0,0 +1,86 @@
+#!/usr/bin/env python3
+import json
+import os
+from diffusers import UNetUnconditionalModel
+from scripts.convert_ldm_original_checkpoint_to_diffusers import convert_ldm_checkpoint
+from huggingface_hub import hf_hub_download
+import torch
+
+model_id = "fusing/latent-diffusion-celeba-256"
+subfolder = "unet"
+#model_id = "fusing/unet-ldm-dummy"
+#subfolder = None
+
+checkpoint = "diffusion_model.pt"
+config = "config.json"
+
+if subfolder is not None:
+    checkpoint = os.path.join(subfolder, checkpoint)
+    config = os.path.join(subfolder, config)
+
+original_checkpoint = torch.load(hf_hub_download(model_id, checkpoint))
+config_path = hf_hub_download(model_id, config)
+
+with open(config_path) as f:
+    config = json.load(f)
+
+checkpoint = convert_ldm_checkpoint(original_checkpoint, config)
+
+
+def current_codebase_conversion():
+    model = UNetUnconditionalModel.from_pretrained(model_id, subfolder=subfolder, ldm=True)
+    model.eval()
+
+    torch.manual_seed(0)
+    if torch.cuda.is_available():
+        torch.cuda.manual_seed_all(0)
+
+    noise = torch.randn(1, model.config.in_channels, model.config.image_size, model.config.image_size)
+    time_step = torch.tensor([10] * noise.shape[0])
+
+    with torch.no_grad():
+        output = model(noise, time_step)
+
+    return model.state_dict()
+
+
+currently_converted_checkpoint = current_codebase_conversion()
+torch.save(currently_converted_checkpoint, 'currently_converted_checkpoint.pt')
+
+
+def diff_between_checkpoints(ch_0, ch_1):
+    all_layers_included = False
+
+    if not set(ch_0.keys()) == set(ch_1.keys()):
+        print(f"Contained in ch_0 and not in ch_1 (Total: {len((set(ch_0.keys()) - set(ch_1.keys())))})")
+        for key in sorted(list((set(ch_0.keys()) - set(ch_1.keys())))):
+            print(f"\t{key}")
+
+        print(f"Contained in ch_1 and not in ch_0 (Total: {len((set(ch_1.keys()) - set(ch_0.keys())))})")
+        for key in sorted(list((set(ch_1.keys()) - set(ch_0.keys())))):
+            print(f"\t{key}")
+    else:
+        print("Keys are the same between the two checkpoints")
+        all_layers_included = True
+
+    keys = ch_0.keys()
+    non_equal_keys = []
+
+    if all_layers_included:
+        for key in keys:
+            try:
+                if not torch.allclose(ch_0[key].cpu(), ch_1[key].cpu()):
+                    non_equal_keys.append(f'{key}. Diff: {torch.max(torch.abs(ch_0[key].cpu() - ch_1[key].cpu()))}')
+
+            except RuntimeError as e:
+                print(e)
+                non_equal_keys.append(f'{key}. Diff in shape: {ch_0[key].size()} vs {ch_1[key].size()}')
+
+        if len(non_equal_keys):
+            non_equal_keys = '\n\t'.join(non_equal_keys)
+            print(f"These keys do not satisfy equivalence requirement:\n\t{non_equal_keys}")
+        else:
+            print("All keys are equal across checkpoints.")
+
+
+diff_between_checkpoints(currently_converted_checkpoint, checkpoint)

scripts/convert_ldm_original_checkpoint_to_diffusers.py

@@ -72,7 +72,7 @@ def renew_attention_paths(old_list, n_shave_prefix_segments=0):
     return mapping
 
 
-def assign_to_checkpoint(paths, checkpoint, old_checkpoint, attention_paths_to_split=None, additional_replacements=None):
+def assign_to_checkpoint(paths, checkpoint, old_checkpoint, attention_paths_to_split=None, additional_replacements=None, config=None):
     """
     This does the final conversion step: take locally converted weights and apply a global renaming
     to them. It splits attention layers, and takes into account additional replacements
@@ -85,11 +85,19 @@ def assign_to_checkpoint(paths, checkpoint, old_checkpoint, attention_paths_to_s
     # Splits the attention layers into three variables.
     if attention_paths_to_split is not None:
         for path, path_map in attention_paths_to_split.items():
-            query, key, value = torch.split(old_checkpoint[path], int(old_checkpoint[path].shape[0] / 3))
+            old_tensor = old_checkpoint[path]
+            channels = old_tensor.shape[0] // 3
 
-            checkpoint[path_map['query']] = query
+            target_shape = (-1, channels) if len(old_tensor.shape) == 3 else (-1)
-            checkpoint[path_map['key']] = key
+
-            checkpoint[path_map['value']] = value
+            num_heads = old_tensor.shape[0] // config["num_head_channels"] // 3
+
+            old_tensor = old_tensor.reshape((num_heads, 3 * channels // num_heads) + old_tensor.shape[1:])
+            query, key, value = old_tensor.split(channels // num_heads, dim=1)
+
+            checkpoint[path_map['query']] = query.reshape(target_shape)
+            checkpoint[path_map['key']] = key.reshape(target_shape)
+            checkpoint[path_map['value']] = value.reshape(target_shape)
 
     for path in paths:
         new_path = path['new']
@@ -107,7 +115,11 @@ def assign_to_checkpoint(paths, checkpoint, old_checkpoint, attention_paths_to_s
             for replacement in additional_replacements:
                 new_path = new_path.replace(replacement['old'], replacement['new'])
 
-        checkpoint[new_path] = old_checkpoint[path['old']]
+        # proj_attn.weight has to be converted from conv 1D to linear
+        if "proj_attn.weight" in new_path:
+            checkpoint[new_path] = old_checkpoint[path['old']][:, :, 0]
+        else:
+            checkpoint[new_path] = old_checkpoint[path['old']]
 
 
 def convert_ldm_checkpoint(checkpoint, config):
@@ -155,7 +167,7 @@ def convert_ldm_checkpoint(checkpoint, config):
         paths = renew_resnet_paths(resnets)
         meta_path = {'old': f'input_blocks.{i}.0', 'new': f'downsample_blocks.{block_id}.resnets.{layer_in_block_id}'}
         resnet_op = {'old': 'resnets.2.op', 'new': 'downsamplers.0.op'}
-        assign_to_checkpoint(paths, new_checkpoint, checkpoint, additional_replacements=[meta_path, resnet_op])
+        assign_to_checkpoint(paths, new_checkpoint, checkpoint, additional_replacements=[meta_path, resnet_op], config=config)
 
         if len(attentions):
             paths = renew_attention_paths(attentions)
@@ -177,19 +189,19 @@ def convert_ldm_checkpoint(checkpoint, config):
                 new_checkpoint,
                 checkpoint,
                 additional_replacements=[meta_path],
-                attention_paths_to_split=to_split
+                attention_paths_to_split=to_split,
+                config=config
             )
 
-
     resnet_0 = middle_blocks[0]
     attentions = middle_blocks[1]
     resnet_1 = middle_blocks[2]
 
     resnet_0_paths = renew_resnet_paths(resnet_0)
-    assign_to_checkpoint(resnet_0_paths, new_checkpoint, checkpoint)
+    assign_to_checkpoint(resnet_0_paths, new_checkpoint, checkpoint, config=config)
 
     resnet_1_paths = renew_resnet_paths(resnet_1)
-    assign_to_checkpoint(resnet_1_paths, new_checkpoint, checkpoint)
+    assign_to_checkpoint(resnet_1_paths, new_checkpoint, checkpoint, config=config)
 
     attentions_paths = renew_attention_paths(attentions)
     to_split = {
@@ -204,7 +216,7 @@ def convert_ldm_checkpoint(checkpoint, config):
             'value': 'mid.attentions.0.value.weight',
         },
     }
-    assign_to_checkpoint(attentions_paths, new_checkpoint, checkpoint, attention_paths_to_split=to_split)
+    assign_to_checkpoint(attentions_paths, new_checkpoint, checkpoint, attention_paths_to_split=to_split, config=config)
 
     for i in range(num_output_blocks):
         block_id = i // (config['num_res_blocks'] + 1)
@@ -227,7 +239,7 @@ def convert_ldm_checkpoint(checkpoint, config):
             paths = renew_resnet_paths(resnets)
 
             meta_path = {'old': f'output_blocks.{i}.0', 'new': f'upsample_blocks.{block_id}.resnets.{layer_in_block_id}'}
-            assign_to_checkpoint(paths, new_checkpoint, checkpoint, additional_replacements=[meta_path])
+            assign_to_checkpoint(paths, new_checkpoint, checkpoint, additional_replacements=[meta_path], config=config)
 
             if ['conv.weight', 'conv.bias'] in output_block_list.values():
                 index = list(output_block_list.values()).index(['conv.weight', 'conv.bias'])
@@ -238,7 +250,6 @@ def convert_ldm_checkpoint(checkpoint, config):
                 if len(attentions) == 2:
                     attentions = []
 
-
             if len(attentions):
                 paths = renew_attention_paths(attentions)
                 meta_path = {
@@ -262,7 +273,8 @@ def convert_ldm_checkpoint(checkpoint, config):
                     new_checkpoint,
                     checkpoint,
                     additional_replacements=[meta_path],
-                    attention_paths_to_split=to_split if any('qkv' in key for key in attentions) else None
+                    attention_paths_to_split=to_split if any('qkv' in key for key in attentions) else None,
+                    config=config,
                 )
         else:
             resnet_0_paths = renew_resnet_paths(output_block_layers, n_shave_prefix_segments=1)
@@ -296,7 +308,6 @@ if __name__ == "__main__":
 
     args = parser.parse_args()
 
-
     checkpoint = torch.load(args.checkpoint_path)
 
     with open(args.config_file) as f:
@@ -304,6 +315,3 @@ if __name__ == "__main__":
 
     converted_checkpoint = convert_ldm_checkpoint(checkpoint, config)
     torch.save(checkpoint, args.dump_path)
-
-
-