From 5989a85edb7fdf24a41ee65cce23403178cc6dd9 Mon Sep 17 00:00:00 2001
From: Steven Liu <59462357+stevhliu@users.noreply.github.com>
Date: Thu, 3 Aug 2023 11:03:42 -0700
Subject: [PATCH] [docs] Distilled SD (#4442)
* first draft
* add blog link
---
docs/source/en/_toctree.yml | 2 +
.../source/en/using-diffusers/distilled_sd.md | 121 ++++++++++++++++++
2 files changed, 123 insertions(+)
create mode 100644 docs/source/en/using-diffusers/distilled_sd.md
diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml
index 795b023809..ac3e02a27f 100644
--- a/docs/source/en/_toctree.yml
+++ b/docs/source/en/_toctree.yml
@@ -50,6 +50,8 @@
title: Textual inversion
- local: training/distributed_inference
title: Distributed inference with multiple GPUs
+ - local: using-diffusers/distilled_sd
+ title: Distilled Stable Diffusion inference
- local: using-diffusers/reusing_seeds
title: Improve image quality with deterministic generation
- local: using-diffusers/control_brightness
diff --git a/docs/source/en/using-diffusers/distilled_sd.md b/docs/source/en/using-diffusers/distilled_sd.md
new file mode 100644
index 0000000000..7653300b92
--- /dev/null
+++ b/docs/source/en/using-diffusers/distilled_sd.md
@@ -0,0 +1,121 @@
+# Distilled Stable Diffusion inference
+
+[[open-in-colab]]
+
+Stable Diffusion inference can be a computationally intensive process because it must iteratively denoise the latents to generate an image. To reduce the computational burden, you can use a *distilled* version of the Stable Diffusion model from [Nota AI](https://huggingface.co/nota-ai). The distilled version of their Stable Diffusion model eliminates some of the residual and attention blocks from the UNet, reducing the model size by 51% and improving latency on CPU/GPU by 43%.
+
+
+
+Read this [blog post](https://huggingface.co/blog/sd_distillation) to learn more about how knowledge distillation training works to produce a faster, smaller, and cheaper generative model.
+
+
+
+Let's load the distilled Stable Diffusion model and compare it against the original Stable Diffusion model:
+
+```py
+from diffusers import StableDiffusionPipeline
+import torch
+
+distilled = StableDiffusionPipeline.from_pretrained(
+ "nota-ai/bk-sdm-small", torch_dtype=torch.float16, use_safetensors=True,
+).to("cuda")
+
+original = StableDiffusionPipeline.from_pretrained(
+ "CompVis/stable-diffusion-v1-4", torch_dtype=torch.float16, use_safetensors=True,
+).to("cuda")
+```
+
+Given a prompt, get the inference time for the original model:
+
+```py
+import time
+
+seed = 2023
+generator = torch.manual_seed(seed)
+
+NUM_ITERS_TO_RUN = 3
+NUM_INFERENCE_STEPS = 25
+NUM_IMAGES_PER_PROMPT = 4
+
+prompt = "a golden vase with different flowers"
+
+start = time.time_ns()
+for _ in range(NUM_ITERS_TO_RUN):
+ images = original(
+ prompt,
+ num_inference_steps=NUM_INFERENCE_STEPS,
+ generator=generator,
+ num_images_per_prompt=NUM_IMAGES_PER_PROMPT
+ ).images
+end = time.time_ns()
+original_sd = f"{(end - start) / 1e6:.1f}"
+
+print(f"Execution time -- {original_sd} ms\n")
+"Execution time -- 45781.5 ms"
+```
+
+Time the distilled model inference:
+
+```py
+start = time.time_ns()
+for _ in range(NUM_ITERS_TO_RUN):
+ images = distilled(
+ prompt,
+ num_inference_steps=NUM_INFERENCE_STEPS,
+ generator=generator,
+ num_images_per_prompt=NUM_IMAGES_PER_PROMPT
+ ).images
+end = time.time_ns()
+
+distilled_sd = f"{(end - start) / 1e6:.1f}"
+print(f"Execution time -- {distilled_sd} ms\n")
+"Execution time -- 29884.2 ms"
+```
+
+
+
+
+
original Stable Diffusion (45781.5 ms)
+
+
+
+
distilled Stable Diffusion (29884.2 ms)
+
+
+
+
+
distilled Stable Diffusion + Tiny AutoEncoder (27165.7 ms)
+
+