dreambooth if docs - stage II, more info (huggingface#3628)

* dreambooth if docs - stage II, more info * Update docs/source/en/training/dreambooth.mdx Co-authored-by: Patrick von Platen <[email protected]> * Update docs/source/en/training/dreambooth.mdx Co-authored-by: Patrick von Platen <[email protected]> * Update docs/source/en/training/dreambooth.mdx Co-authored-by: Sayak Paul <[email protected]> * download instructions for downsized images * update source README to match docs --------- Co-authored-by: Patrick von Platen <[email protected]> Co-authored-by: Sayak Paul <[email protected]>
cardinalblue · Jun 2, 2023 · 5911a3a · 5911a3a
1 parent b7af946
commit 5911a3a
Show file tree

Hide file tree

Showing 2 changed files with 295 additions and 31 deletions.
diff --git a/docs/source/en/training/dreambooth.mdx b/docs/source/en/training/dreambooth.mdx
@@ -502,9 +502,65 @@ You may also run inference from any of the [saved training checkpoints](#inferen
 
 ## IF
 
-You can use the lora and full dreambooth scripts to also train the text to image [IF model](https://huggingface.co/DeepFloyd/IF-I-XL-v1.0). A few alternative cli flags are needed due to the model size, the expected input resolution, and the text encoder conventions.
+You can use the lora and full dreambooth scripts to train the text to image [IF model](https://huggingface.co/DeepFloyd/IF-I-XL-v1.0) and the stage II upscaler 
+[IF model](https://huggingface.co/DeepFloyd/IF-II-L-v1.0).
 
-### LoRA Dreambooth
+Note that IF has a predicted variance, and our finetuning scripts only train the models predicted error, so for finetuned IF models we switch to a fixed
+variance schedule. The full finetuning scripts will update the scheduler config for the full saved model. However, when loading saved LoRA weights, you
+must also update the pipeline's scheduler config.
+
+```py
+from diffusers import DiffusionPipeline
+
+pipe = DiffusionPipeline.from_pretrained("DeepFloyd/IF-I-XL-v1.0")
+
+pipe.load_lora_weights("<lora weights path>")
+
+# Update scheduler config to fixed variance schedule
+pipe.scheduler = pipe.scheduler.__class__.from_config(pipe.scheduler.config, variance_type="fixed_small")
+```
+
+Additionally, a few alternative cli flags are needed for IF.
+
+`--resolution=64`: IF is a pixel space diffusion model. In order to operate on un-compressed pixels, the input images are of a much smaller resolution. 
+
+`--pre_compute_text_embeddings`: IF uses [T5](https://huggingface.co/docs/transformers/model_doc/t5) for its text encoder. In order to save GPU memory, we pre compute all text embeddings and then de-allocate
+T5.
+
+`--tokenizer_max_length=77`: T5 has a longer default text length, but the default IF encoding procedure uses a smaller number.
+
+`--text_encoder_use_attention_mask`: T5 passes the attention mask to the text encoder.
+
+### Tips and Tricks
+We find LoRA to be sufficient for finetuning the stage I model as the low resolution of the model makes representing finegrained detail hard regardless.
+
+For common and/or not-visually complex object concepts, you can get away with not-finetuning the upscaler. Just be sure to adjust the prompt passed to the
+upscaler to remove the new token from the instance prompt. I.e. if your stage I prompt is "a sks dog", use "a dog" for your stage II prompt.
+
+For finegrained detail like faces that aren't present in the original training set, we find that full finetuning of the stage II upscaler is better than 
+LoRA finetuning stage II.
+
+For finegrained detail like faces, we find that lower learning rates work best.
+
+For stage II, we find that lower learning rates are also needed.
+
+### Stage II additional validation images
+
+The stage II validation requires images to upscale, we can download a downsized version of the training set:
+
+```py
+from huggingface_hub import snapshot_download
+
+local_dir = "./dog_downsized"
+snapshot_download(
+    "diffusers/dog-example-downsized",
+    local_dir=local_dir,
+    repo_type="dataset",
+    ignore_patterns=".gitattributes",
+)
+```
+
+### IF stage I LoRA Dreambooth
 This training configuration requires ~28 GB VRAM.
 
 ```sh
@@ -518,7 +574,7 @@ accelerate launch train_dreambooth_lora.py \
   --instance_data_dir=$INSTANCE_DIR \
   --output_dir=$OUTPUT_DIR \
   --instance_prompt="a sks dog" \
-  --resolution=64 \ # The input resolution of the IF unet is 64x64
+  --resolution=64 \
   --train_batch_size=4 \
   --gradient_accumulation_steps=1 \
   --learning_rate=5e-6 \
@@ -527,16 +583,57 @@ accelerate launch train_dreambooth_lora.py \
   --validation_prompt="a sks dog" \
   --validation_epochs=25 \
   --checkpointing_steps=100 \
-  --pre_compute_text_embeddings \ # Pre compute text embeddings to that T5 doesn't have to be kept in memory
-  --tokenizer_max_length=77 \ # IF expects an override of the max token length
-  --text_encoder_use_attention_mask # IF expects attention mask for text embeddings
+  --pre_compute_text_embeddings \
+  --tokenizer_max_length=77 \
+  --text_encoder_use_attention_mask
 ```
 
-### Full Dreambooth
-Due to the size of the optimizer states, we recommend training the full XL IF model with 8bit adam. 
-Using 8bit adam and the rest of the following config, the model can be trained in ~48 GB VRAM.
+### IF stage II LoRA Dreambooth
 
-For full dreambooth, IF requires very low learning rates. With higher learning rates model quality will degrade.
+`--validation_images`: These images are upscaled during validation steps.
+
+`--class_labels_conditioning=timesteps`: Pass additional conditioning to the UNet needed for stage II.
+
+`--learning_rate=1e-6`: Lower learning rate than stage I.
+
+`--resolution=256`: The upscaler expects higher resolution inputs
+
+```sh
+export MODEL_NAME="DeepFloyd/IF-II-L-v1.0"
+export INSTANCE_DIR="dog"
+export OUTPUT_DIR="dreambooth_dog_upscale"
+export VALIDATION_IMAGES="dog_downsized/image_1.png dog_downsized/image_2.png dog_downsized/image_3.png dog_downsized/image_4.png"
+
+python train_dreambooth_lora.py \
+    --report_to wandb \
+    --pretrained_model_name_or_path=$MODEL_NAME \
+    --instance_data_dir=$INSTANCE_DIR \
+    --output_dir=$OUTPUT_DIR \
+    --instance_prompt="a sks dog" \
+    --resolution=256 \
+    --train_batch_size=4 \
+    --gradient_accumulation_steps=1 \
+    --learning_rate=1e-6 \ 
+    --max_train_steps=2000 \
+    --validation_prompt="a sks dog" \
+    --validation_epochs=100 \
+    --checkpointing_steps=500 \
+    --pre_compute_text_embeddings \
+    --tokenizer_max_length=77 \
+    --text_encoder_use_attention_mask \
+    --validation_images $VALIDATION_IMAGES \
+    --class_labels_conditioning=timesteps
+```
+
+### IF Stage I Full Dreambooth
+`--skip_save_text_encoder`: When training the full model, this will skip saving the entire T5 with the finetuned model. You can still load the pipeline
+with a T5 loaded from the original model.
+
+`use_8bit_adam`: Due to the size of the optimizer states, we recommend training the full XL IF model with 8bit adam. 
+
+`--learning_rate=1e-7`: For full dreambooth, IF requires very low learning rates. With higher learning rates model quality will degrade.
+
+Using 8bit adam and a batch size of 4, the model can be trained in ~48 GB VRAM.
 
 ```sh
 export MODEL_NAME="DeepFloyd/IF-I-XL-v1.0"
@@ -549,17 +646,52 @@ accelerate launch train_dreambooth.py \
   --instance_data_dir=$INSTANCE_DIR \
   --output_dir=$OUTPUT_DIR \
   --instance_prompt="a photo of sks dog" \
-  --resolution=64 \ # The input resolution of the IF unet is 64x64
+  --resolution=64 \
   --train_batch_size=4 \
   --gradient_accumulation_steps=1 \
   --learning_rate=1e-7 \
   --max_train_steps=150 \
   --validation_prompt "a photo of sks dog" \
   --validation_steps 25 \
-  --text_encoder_use_attention_mask \ # IF expects attention mask for text embeddings
-  --tokenizer_max_length 77 \ # IF expects an override of the max token length
-  --pre_compute_text_embeddings \ # Pre compute text embeddings to that T5 doesn't have to be kept in memory
+  --text_encoder_use_attention_mask \
+  --tokenizer_max_length 77 \
+  --pre_compute_text_embeddings \
   --use_8bit_adam \ # 
   --set_grads_to_none \
-  --skip_save_text_encoder # do not save the full T5 text encoder with the model
-```
+  --skip_save_text_encoder \
+  --push_to_hub
+```
+
+### IF Stage II Full Dreambooth
+
+`--learning_rate=1e-8`: Even lower learning rate.
+
+`--resolution=256`: The upscaler expects higher resolution inputs
+
+```sh
+export MODEL_NAME="DeepFloyd/IF-II-L-v1.0"
+export INSTANCE_DIR="dog"
+export OUTPUT_DIR="dreambooth_dog_upscale"
+export VALIDATION_IMAGES="dog_downsized/image_1.png dog_downsized/image_2.png dog_downsized/image_3.png dog_downsized/image_4.png"
+
+accelerate launch train_dreambooth.py \
+  --report_to wandb \
+  --pretrained_model_name_or_path=$MODEL_NAME \
+  --instance_data_dir=$INSTANCE_DIR \
+  --output_dir=$OUTPUT_DIR \
+  --instance_prompt="a sks dog" \
+  --resolution=256 \
+  --train_batch_size=2 \
+  --gradient_accumulation_steps=2 \
+  --learning_rate=1e-8 \
+  --max_train_steps=2000 \
+  --validation_prompt="a sks dog" \
+  --validation_steps=150 \
+  --checkpointing_steps=500 \
+  --pre_compute_text_embeddings \
+  --tokenizer_max_length=77 \
+  --text_encoder_use_attention_mask \
+  --validation_images $VALIDATION_IMAGES \
+  --class_labels_conditioning timesteps \
+  --push_to_hub
+```