Removing autocast for 35-25% speedup.

Author: Narsil

README.md

@@ -76,15 +76,13 @@ You need to accept the model license before downloading or using the Stable Diff
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline
 
 pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
 **Note**: If you don't want to use the token, you can also simply download the model weights
@@ -104,8 +102,7 @@ pipe = StableDiffusionPipeline.from_pretrained("./stable-diffusion-v1-4")
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
 If you are limited by GPU memory, you might want to consider using the model in `fp16` as 
@@ -123,8 +120,7 @@ pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
 pipe.enable_attention_slicing()
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 ```
 
 Finally, if you wish to use a different scheduler, you can simply instantiate
@@ -149,8 +145,7 @@ pipe = StableDiffusionPipeline.from_pretrained(
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -160,7 +155,6 @@ image.save("astronaut_rides_horse.png")
 The `StableDiffusionImg2ImgPipeline` lets you pass a text prompt and an initial image to condition the generation of new images.
 
 ```python
-from torch import autocast
 import requests
 import torch
 from PIL import Image
@@ -190,8 +184,7 @@ init_image = init_image.resize((768, 512))
 
 prompt = "A fantasy landscape, trending on artstation"
 
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
+images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
 
 images[0].save("fantasy_landscape.png")
 ```
@@ -204,7 +197,6 @@ The `StableDiffusionInpaintPipeline` lets you edit specific parts of an image by
 ```python
 from io import BytesIO
 
-from torch import autocast
 import torch
 import requests
 import PIL
@@ -234,8 +226,7 @@ pipe = StableDiffusionInpaintPipeline.from_pretrained(
 pipe = pipe.to(device)
 
 prompt = "a cat sitting on a bench"
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
+images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
 
 images[0].save("cat_on_bench.png")
 ```
@@ -258,7 +249,6 @@ If you want to run the code yourself 💻, you can try out:
 - [Text-to-Image Latent Diffusion](https://huggingface.co/CompVis/ldm-text2im-large-256)
 ```python
 # !pip install diffusers transformers
-from torch import autocast
 from diffusers import DiffusionPipeline
 
 device = "cuda"
@@ -270,16 +260,14 @@ ldm = ldm.to(device)
 
 # run pipeline in inference (sample random noise and denoise)
 prompt = "A painting of a squirrel eating a burger"
-with autocast(device):
-    image = ldm([prompt], num_inference_steps=50, eta=0.3, guidance_scale=6).images[0]
+image = ldm([prompt], num_inference_steps=50, eta=0.3, guidance_scale=6).images[0]
 
 # save image
 image.save("squirrel.png")
 ```
 - [Unconditional Diffusion with discrete scheduler](https://huggingface.co/google/ddpm-celebahq-256)
 ```python
 # !pip install diffusers
-from torch import autocast
 from diffusers import DDPMPipeline, DDIMPipeline, PNDMPipeline
 
 model_id = "google/ddpm-celebahq-256"
@@ -290,8 +278,7 @@ ddpm = DDPMPipeline.from_pretrained(model_id)  # you can replace DDPMPipeline wi
 ddpm.to(device)
 
 # run pipeline in inference (sample random noise and denoise)
-with autocast("cuda"):
-    image = ddpm().images[0]
+image = ddpm().images[0]
 
 # save image
 image.save("ddpm_generated_image.png")

src/diffusers/models/resnet.py

@@ -331,9 +331,7 @@ def __init__(
     def forward(self, x, temb):
         hidden_states = x
 
-        # make sure hidden states is in float32
-        # when running in half-precision
-        hidden_states = self.norm1(hidden_states.float()).type(hidden_states.dtype)
+        hidden_states = self.norm1(hidden_states).type(hidden_states.dtype)
         hidden_states = self.nonlinearity(hidden_states)
 
         if self.upsample is not None:
@@ -349,9 +347,7 @@ def forward(self, x, temb):
             temb = self.time_emb_proj(self.nonlinearity(temb))[:, :, None, None]
             hidden_states = hidden_states + temb
 
-        # make sure hidden states is in float32
-        # when running in half-precision
-        hidden_states = self.norm2(hidden_states.float()).type(hidden_states.dtype)
+        hidden_states = self.norm2(hidden_states).type(hidden_states.dtype)
         hidden_states = self.nonlinearity(hidden_states)
 
         hidden_states = self.dropout(hidden_states)

src/diffusers/models/unet_2d_condition.py

@@ -222,6 +222,11 @@ def forward(
         timesteps = timesteps.expand(sample.shape[0])
 
         t_emb = self.time_proj(timesteps)
+
+        # timesteps does not contain any weights and will always return f32 tensors
+        # but time_embedding might actually be running in fp16. so we need to cast here.
+        # there might be better ways to encapsulate this.
+        t_emb = t_emb.to(dtype=sample.dtype)
         emb = self.time_embedding(t_emb)
 
         # 2. pre-process
@@ -258,9 +263,7 @@ def forward(
                 sample = upsample_block(hidden_states=sample, temb=emb, res_hidden_states_tuple=res_samples)
 
         # 6. post-process
-        # make sure hidden states is in float32
-        # when running in half-precision
-        sample = self.conv_norm_out(sample.float()).type(sample.dtype)
+        sample = self.conv_norm_out(sample).type(sample.dtype)
         sample = self.conv_act(sample)
         sample = self.conv_out(sample)
 

src/diffusers/pipelines/README.md

@@ -86,15 +86,13 @@ logic including pre-processing, an unrolled diffusion loop, and post-processing
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline, LMSDiscreteScheduler
 
 pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).images[0]  
+image = pipe(prompt).images[0]  
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -104,33 +102,7 @@ image.save("astronaut_rides_horse.png")
 The `StableDiffusionImg2ImgPipeline` lets you pass a text prompt and an initial image to condition the generation of new images.
 
 ```python
-from torch import autocast
-import requests
-from PIL import Image
-from io import BytesIO
-
-from diffusers import StableDiffusionImg2ImgPipeline
-
-# load the pipeline
-device = "cuda"
-pipe = StableDiffusionImg2ImgPipeline.from_pretrained(
-    "CompVis/stable-diffusion-v1-4",
-    revision="fp16", 
-    torch_dtype=torch.float16,
-    use_auth_token=True
-).to(device)
-
-# let's download an initial image
-url = "https://raw.githubusercontent.com/CompVis/stable-diffusion/main/assets/stable-samples/img2img/sketch-mountains-input.jpg"
-
-response = requests.get(url)
-init_image = Image.open(BytesIO(response.content)).convert("RGB")
-init_image = init_image.resize((768, 512))
-
-prompt = "A fantasy landscape, trending on artstation"
-
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
+images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5).images
 
 images[0].save("fantasy_landscape.png")
 ```
@@ -148,7 +120,6 @@ The `StableDiffusionInpaintPipeline` lets you edit specific parts of an image by
 ```python
 from io import BytesIO
 
-from torch import autocast
 import requests
 import PIL
 
@@ -173,8 +144,7 @@ pipe = StableDiffusionInpaintPipeline.from_pretrained(
 ).to(device)
 
 prompt = "a cat sitting on a bench"
-with autocast("cuda"):
-    images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
+images = pipe(prompt=prompt, init_image=init_image, mask_image=mask_image, strength=0.75).images
 
 images[0].save("cat_on_bench.png")
 ```

src/diffusers/pipelines/stable_diffusion/README.md

@@ -59,15 +59,13 @@ pipe = StableDiffusionPipeline.from_pretrained("./stable-diffusion-v1-4")
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline
 
 pipe = StableDiffusionPipeline.from_pretrained("CompVis/stable-diffusion-v1-4", use_auth_token=True)
 pipe = pipe.to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).sample[0]  
+image = pipe(prompt).sample[0]  
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -76,7 +74,6 @@ image.save("astronaut_rides_horse.png")
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline, DDIMScheduler
 
 scheduler = DDIMScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", clip_sample=False, set_alpha_to_one=False)
@@ -88,8 +85,7 @@ pipe = StableDiffusionPipeline.from_pretrained(
 ).to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).sample[0]  
+image = pipe(prompt).sample[0]  
 
 image.save("astronaut_rides_horse.png")
 ```
@@ -98,7 +94,6 @@ image.save("astronaut_rides_horse.png")
 
 ```python
 # make sure you're logged in with `huggingface-cli login`
-from torch import autocast
 from diffusers import StableDiffusionPipeline, LMSDiscreteScheduler
 
 lms = LMSDiscreteScheduler(
@@ -114,8 +109,7 @@ pipe = StableDiffusionPipeline.from_pretrained(
 ).to("cuda")
 
 prompt = "a photo of an astronaut riding a horse on mars"
-with autocast("cuda"):
-    image = pipe(prompt).sample[0]  
+image = pipe(prompt).sample[0]  
 
 image.save("astronaut_rides_horse.png")
 ```

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -205,16 +205,18 @@ def __call__(
         # However this currently doesn't work in `mps`.
         latents_device = "cpu" if self.device.type == "mps" else self.device
         latents_shape = (batch_size, self.unet.in_channels, height // 8, width // 8)
+        latents_dtype = text_embeddings.dtype
         if latents is None:
             latents = torch.randn(
                 latents_shape,
                 generator=generator,
                 device=latents_device,
+                dtype=latents_dtype
             )
         else:
             if latents.shape != latents_shape:
                 raise ValueError(f"Unexpected latents shape, got {latents.shape}, expected {latents_shape}")
-        latents = latents.to(self.device)
+            latents = latents.to(self.device)
 
         # set timesteps
         accepts_offset = "offset" in set(inspect.signature(self.scheduler.set_timesteps).parameters.keys())
@@ -268,6 +270,12 @@ def __call__(
 
         # run safety checker
         safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
+
+        # XXX: it might be better to check against the actual dtype of the safety checker since
+        # it might want to run in a different precision, but the safety checker does not expose
+        # a `dtype` /`precision` itself, so this is a good enough proxy for running pipelines in
+        # both f16 / f32
+        safety_cheker_input.pixel_values = safety_cheker_input.pixel_values.to(dtype=latents_dtype)
         image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)
 
         if output_type == "pil":