[advanced dreambooth lora] add clip_skip arg (#8715)

* add clip_skip

* style

* smol fix

---------

Co-authored-by: Sayak Paul <[email protected]>

Author: linoytsaban

examples/advanced_diffusion_training/train_dreambooth_lora_sdxl_advanced.py

@@ -573,6 +573,13 @@ def parse_args(input_args=None):
         default=1e-4,
         help="Initial learning rate (after the potential warmup period) to use.",
     )
+    parser.add_argument(
+        "--clip_skip",
+        type=int,
+        default=None,
+        help="Number of layers to be skipped from CLIP while computing the prompt embeddings. A value of 1 means that "
+        "the output of the pre-final layer will be used for computing the prompt embeddings.",
+    )
 
     parser.add_argument(
         "--text_encoder_lr",
@@ -1236,7 +1243,7 @@ def tokenize_prompt(tokenizer, prompt, add_special_tokens=False):
 
 
 # Adapted from pipelines.StableDiffusionXLPipeline.encode_prompt
-def encode_prompt(text_encoders, tokenizers, prompt, text_input_ids_list=None):
+def encode_prompt(text_encoders, tokenizers, prompt, text_input_ids_list=None, clip_skip=None):
     prompt_embeds_list = []
 
     for i, text_encoder in enumerate(text_encoders):
@@ -1253,7 +1260,11 @@ def encode_prompt(text_encoders, tokenizers, prompt, text_input_ids_list=None):
 
         # We are only ALWAYS interested in the pooled output of the final text encoder
         pooled_prompt_embeds = prompt_embeds[0]
-        prompt_embeds = prompt_embeds[-1][-2]
+        if clip_skip is None:
+            prompt_embeds = prompt_embeds[-1][-2]
+        else:
+            # "2" because SDXL always indexes from the penultimate layer.
+            prompt_embeds = prompt_embeds[-1][-(clip_skip + 2)]
         bs_embed, seq_len, _ = prompt_embeds.shape
         prompt_embeds = prompt_embeds.view(bs_embed, seq_len, -1)
         prompt_embeds_list.append(prompt_embeds)
@@ -1830,9 +1841,9 @@ def compute_time_ids(crops_coords_top_left, original_size=None):
         tokenizers = [tokenizer_one, tokenizer_two]
         text_encoders = [text_encoder_one, text_encoder_two]
 
-        def compute_text_embeddings(prompt, text_encoders, tokenizers):
+        def compute_text_embeddings(prompt, text_encoders, tokenizers, clip_skip):
             with torch.no_grad():
-                prompt_embeds, pooled_prompt_embeds = encode_prompt(text_encoders, tokenizers, prompt)
+                prompt_embeds, pooled_prompt_embeds = encode_prompt(text_encoders, tokenizers, prompt, clip_skip)
                 prompt_embeds = prompt_embeds.to(accelerator.device)
                 pooled_prompt_embeds = pooled_prompt_embeds.to(accelerator.device)
             return prompt_embeds, pooled_prompt_embeds
@@ -1842,7 +1853,7 @@ def compute_text_embeddings(prompt, text_encoders, tokenizers):
     # the redundant encoding.
     if freeze_text_encoder and not train_dataset.custom_instance_prompts:
         instance_prompt_hidden_states, instance_pooled_prompt_embeds = compute_text_embeddings(
-            args.instance_prompt, text_encoders, tokenizers
+            args.instance_prompt, text_encoders, tokenizers, args.clip_skip
         )
 
     # Handle class prompt for prior-preservation.
@@ -2052,7 +2063,7 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                 if train_dataset.custom_instance_prompts:
                     if freeze_text_encoder:
                         prompt_embeds, unet_add_text_embeds = compute_text_embeddings(
-                            prompts, text_encoders, tokenizers
+                            prompts, text_encoders, tokenizers, args.clip_skip
                         )
 
                     else:
@@ -2147,6 +2158,7 @@ def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                         tokenizers=None,
                         prompt=None,
                         text_input_ids_list=[tokens_one, tokens_two],
+                        clip_skip=args.clip_skip,
                     )
                     unet_added_conditions.update(
                         {"text_embeds": pooled_prompt_embeds.repeat(elems_to_repeat_text_embeds, 1)}