From f8740dacf36a2b01b8e5e2cce6340790952db797 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:25:00 +0300 Subject: [PATCH 1/9] Revert "[`Docs`] Update and make improvements (#5819)" This reverts commit c697f524761abd2314c030221a3ad2f7791eab4e. --- README.md | 10 ++++++---- docs/source/en/optimization/memory.md | 6 +++--- docs/source/en/tutorials/basic_training.md | 4 ++-- docs/source/en/using-diffusers/write_own_pipeline.md | 4 ++-- docs/source/ko/optimization/fp16.md | 6 +++--- docs/source/ko/tutorials/basic_training.md | 4 ++-- docs/source/ko/using-diffusers/write_own_pipeline.md | 4 ++-- 7 files changed, 20 insertions(+), 18 deletions(-) diff --git a/README.md b/README.md index 489e0d154af2..9655220a2491 100644 --- a/README.md +++ b/README.md @@ -47,7 +47,7 @@ limitations under the License. ## Installation -We recommend installing ๐Ÿค— Diffusers in a virtual environment from PyPI or Conda. For more details about installing [PyTorch](https://pytorch.org/get-started/locally/) and [Flax](https://flax.readthedocs.io/en/latest/#installation), please refer to their official documentation. +We recommend installing ๐Ÿค— Diffusers in a virtual environment from PyPi or Conda. For more details about installing [PyTorch](https://pytorch.org/get-started/locally/) and [Flax](https://flax.readthedocs.io/en/latest/#installation), please refer to their official documentation. ### PyTorch @@ -77,7 +77,7 @@ Please refer to the [How to use Stable Diffusion in Apple Silicon](https://huggi ## Quickstart -Generating outputs is super easy with ๐Ÿค— Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 15000+ checkpoints): +Generating outputs is super easy with ๐Ÿค— Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 14000+ checkpoints): ```python from diffusers import DiffusionPipeline @@ -94,13 +94,14 @@ You can also dig into the models and schedulers toolbox to build your own diffus from diffusers import DDPMScheduler, UNet2DModel from PIL import Image import torch +import numpy as np scheduler = DDPMScheduler.from_pretrained("google/ddpm-cat-256") model = UNet2DModel.from_pretrained("google/ddpm-cat-256").to("cuda") scheduler.set_timesteps(50) sample_size = model.config.sample_size -noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") +noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") input = noise for t in scheduler.timesteps: @@ -135,7 +136,8 @@ You can look out for [issues](https://github.com/huggingface/diffusers/issues) y - See [New model/pipeline](https://github.com/huggingface/diffusers/issues?q=is%3Aopen+is%3Aissue+label%3A%22New+pipeline%2Fmodel%22) to contribute exciting new diffusion models / diffusion pipelines - See [New scheduler](https://github.com/huggingface/diffusers/issues?q=is%3Aopen+is%3Aissue+label%3A%22New+scheduler%22) -Also, say ๐Ÿ‘‹ in our public Discord channel Join us on Discord. We discuss the hottest trends about diffusion models, help each other with contributions, personal projects or just hang out โ˜•. +Also, say ๐Ÿ‘‹ in our public Discord channel Join us on Discord. We discuss the hottest trends about diffusion models, help each other with contributions, personal projects or +just hang out โ˜•. ## Popular Tasks & Pipelines diff --git a/docs/source/en/optimization/memory.md b/docs/source/en/optimization/memory.md index 42a1bcea8fb5..281b65df8d8c 100644 --- a/docs/source/en/optimization/memory.md +++ b/docs/source/en/optimization/memory.md @@ -194,9 +194,9 @@ unet_runs_per_experiment = 50 # load inputs def generate_inputs(): - sample = torch.randn((2, 4, 64, 64), device="cuda", dtype=torch.float16) - timestep = torch.rand(1, device="cuda", dtype=torch.float16) * 999 - encoder_hidden_states = torch.randn((2, 77, 768), device="cuda", dtype=torch.float16) + sample = torch.randn(2, 4, 64, 64).half().cuda() + timestep = torch.rand(1).half().cuda() * 999 + encoder_hidden_states = torch.randn(2, 77, 768).half().cuda() return sample, timestep, encoder_hidden_states diff --git a/docs/source/en/tutorials/basic_training.md b/docs/source/en/tutorials/basic_training.md index c9ce315af41f..3b545cdf572e 100644 --- a/docs/source/en/tutorials/basic_training.md +++ b/docs/source/en/tutorials/basic_training.md @@ -321,13 +321,13 @@ Now you can wrap all these components together in a training loop with ๐Ÿค— Acce ... for step, batch in enumerate(train_dataloader): ... clean_images = batch["images"] ... # Sample noise to add to the images -... noise = torch.randn(clean_images.shape, device=clean_images.device) +... noise = torch.randn(clean_images.shape).to(clean_images.device) ... bs = clean_images.shape[0] ... # Sample a random timestep for each image ... timesteps = torch.randint( ... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device -... ) +... ).long() ... # Add noise to the clean images according to the noise magnitude at each timestep ... # (this is the forward diffusion process) diff --git a/docs/source/en/using-diffusers/write_own_pipeline.md b/docs/source/en/using-diffusers/write_own_pipeline.md index 4ca3fe33223b..d3061e36fe63 100644 --- a/docs/source/en/using-diffusers/write_own_pipeline.md +++ b/docs/source/en/using-diffusers/write_own_pipeline.md @@ -71,7 +71,7 @@ tensor([980, 960, 940, 920, 900, 880, 860, 840, 820, 800, 780, 760, 740, 720, >>> import torch >>> sample_size = model.config.sample_size ->>> noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") +>>> noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") ``` 5. Now write a loop to iterate over the timesteps. At each timestep, the model does a [`UNet2DModel.forward`] pass and returns the noisy residual. The scheduler's [`~DDPMScheduler.step`] method takes the noisy residual, timestep, and input and it predicts the image at the previous timestep. This output becomes the next input to the model in the denoising loop, and it'll repeat until it reaches the end of the `timesteps` array. @@ -216,8 +216,8 @@ Next, generate some initial random noise as a starting point for the diffusion p >>> latents = torch.randn( ... (batch_size, unet.config.in_channels, height // 8, width // 8), ... generator=generator, -... device=torch_device, ... ) +>>> latents = latents.to(torch_device) ``` ### Denoise the image diff --git a/docs/source/ko/optimization/fp16.md b/docs/source/ko/optimization/fp16.md index 0f2c487a75ce..30197305540c 100644 --- a/docs/source/ko/optimization/fp16.md +++ b/docs/source/ko/optimization/fp16.md @@ -273,9 +273,9 @@ unet_runs_per_experiment = 50 # ์ž…๋ ฅ ๋ถˆ๋Ÿฌ์˜ค๊ธฐ def generate_inputs(): - sample = torch.randn((2, 4, 64, 64), device="cuda", dtype=torch.float16) - timestep = torch.rand(1, device="cuda", dtype=torch.float16) * 999 - encoder_hidden_states = torch.randn((2, 77, 768), device="cuda", dtype=torch.float16) + sample = torch.randn(2, 4, 64, 64).half().cuda() + timestep = torch.rand(1).half().cuda() * 999 + encoder_hidden_states = torch.randn(2, 77, 768).half().cuda() return sample, timestep, encoder_hidden_states diff --git a/docs/source/ko/tutorials/basic_training.md b/docs/source/ko/tutorials/basic_training.md index df5e74c22ca8..a4e5e2a0c8bb 100644 --- a/docs/source/ko/tutorials/basic_training.md +++ b/docs/source/ko/tutorials/basic_training.md @@ -322,13 +322,13 @@ TensorBoard์— ๋กœ๊น…, ๊ทธ๋ž˜๋””์–ธํŠธ ๋ˆ„์  ๋ฐ ํ˜ผํ•ฉ ์ •๋ฐ€๋„ ํ•™์Šต์„ ์‰ฝ ... for step, batch in enumerate(train_dataloader): ... clean_images = batch["images"] ... # ์ด๋ฏธ์ง€์— ๋”ํ•  ๋…ธ์ด์ฆˆ๋ฅผ ์ƒ˜ํ”Œ๋งํ•ฉ๋‹ˆ๋‹ค. -... noise = torch.randn(clean_images.shape, device=clean_images.device) +... noise = torch.randn(clean_images.shape).to(clean_images.device) ... bs = clean_images.shape[0] ... # ๊ฐ ์ด๋ฏธ์ง€๋ฅผ ์œ„ํ•œ ๋žœ๋คํ•œ ํƒ€์ž„์Šคํ…(timestep)์„ ์ƒ˜ํ”Œ๋งํ•ฉ๋‹ˆ๋‹ค. ... timesteps = torch.randint( ... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device -... ) +... ).long() ... # ๊ฐ ํƒ€์ž„์Šคํ…์˜ ๋…ธ์ด์ฆˆ ํฌ๊ธฐ์— ๋”ฐ๋ผ ๊นจ๋—ํ•œ ์ด๋ฏธ์ง€์— ๋…ธ์ด์ฆˆ๋ฅผ ์ถ”๊ฐ€ํ•ฉ๋‹ˆ๋‹ค. ... # (์ด๋Š” foward diffusion ๊ณผ์ •์ž…๋‹ˆ๋‹ค.) diff --git a/docs/source/ko/using-diffusers/write_own_pipeline.md b/docs/source/ko/using-diffusers/write_own_pipeline.md index 787c8113bf0d..a6469644566c 100644 --- a/docs/source/ko/using-diffusers/write_own_pipeline.md +++ b/docs/source/ko/using-diffusers/write_own_pipeline.md @@ -71,7 +71,7 @@ specific language governing permissions and limitations under the License. >>> import torch >>> sample_size = model.config.sample_size - >>> noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") + >>> noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") ``` 5. ์ด์ œ timestep์„ ๋ฐ˜๋ณตํ•˜๋Š” ๋ฃจํ”„๋ฅผ ์ž‘์„ฑํ•ฉ๋‹ˆ๋‹ค. ๊ฐ timestep์—์„œ ๋ชจ๋ธ์€ [`UNet2DModel.forward`]๋ฅผ ํ†ตํ•ด noisy residual์„ ๋ฐ˜ํ™˜ํ•ฉ๋‹ˆ๋‹ค. ์Šค์ผ€์ค„๋Ÿฌ์˜ [`~DDPMScheduler.step`] ๋ฉ”์„œ๋“œ๋Š” noisy residual, timestep, ๊ทธ๋ฆฌ๊ณ  ์ž…๋ ฅ์„ ๋ฐ›์•„ ์ด์ „ timestep์—์„œ ์ด๋ฏธ์ง€๋ฅผ ์˜ˆ์ธกํ•ฉ๋‹ˆ๋‹ค. ์ด ์ถœ๋ ฅ์€ ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ ๋ฃจํ”„์˜ ๋ชจ๋ธ์— ๋Œ€ํ•œ ๋‹ค์Œ ์ž…๋ ฅ์ด ๋˜๋ฉฐ, `timesteps` ๋ฐฐ์—ด์˜ ๋์— ๋„๋‹ฌํ•  ๋•Œ๊นŒ์ง€ ๋ฐ˜๋ณต๋ฉ๋‹ˆ๋‹ค. @@ -212,8 +212,8 @@ Stable Diffusion ์€ text-to-image *latent diffusion* ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. latent di >>> latents = torch.randn( ... (batch_size, unet.in_channels, height // 8, width // 8), ... generator=generator, -... device=torch_device, ... ) +>>> latents = latents.to(torch_device) ``` ### ์ด๋ฏธ์ง€ ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ From b522cdd1c1e53211559f2dd951c370a96fc71dca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:31:04 +0300 Subject: [PATCH 2/9] Update README.md --- README.md | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/README.md b/README.md index 9655220a2491..489e0d154af2 100644 --- a/README.md +++ b/README.md @@ -47,7 +47,7 @@ limitations under the License. ## Installation -We recommend installing ๐Ÿค— Diffusers in a virtual environment from PyPi or Conda. For more details about installing [PyTorch](https://pytorch.org/get-started/locally/) and [Flax](https://flax.readthedocs.io/en/latest/#installation), please refer to their official documentation. +We recommend installing ๐Ÿค— Diffusers in a virtual environment from PyPI or Conda. For more details about installing [PyTorch](https://pytorch.org/get-started/locally/) and [Flax](https://flax.readthedocs.io/en/latest/#installation), please refer to their official documentation. ### PyTorch @@ -77,7 +77,7 @@ Please refer to the [How to use Stable Diffusion in Apple Silicon](https://huggi ## Quickstart -Generating outputs is super easy with ๐Ÿค— Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 14000+ checkpoints): +Generating outputs is super easy with ๐Ÿค— Diffusers. To generate an image from text, use the `from_pretrained` method to load any pretrained diffusion model (browse the [Hub](https://huggingface.co/models?library=diffusers&sort=downloads) for 15000+ checkpoints): ```python from diffusers import DiffusionPipeline @@ -94,14 +94,13 @@ You can also dig into the models and schedulers toolbox to build your own diffus from diffusers import DDPMScheduler, UNet2DModel from PIL import Image import torch -import numpy as np scheduler = DDPMScheduler.from_pretrained("google/ddpm-cat-256") model = UNet2DModel.from_pretrained("google/ddpm-cat-256").to("cuda") scheduler.set_timesteps(50) sample_size = model.config.sample_size -noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") +noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") input = noise for t in scheduler.timesteps: @@ -136,8 +135,7 @@ You can look out for [issues](https://github.com/huggingface/diffusers/issues) y - See [New model/pipeline](https://github.com/huggingface/diffusers/issues?q=is%3Aopen+is%3Aissue+label%3A%22New+pipeline%2Fmodel%22) to contribute exciting new diffusion models / diffusion pipelines - See [New scheduler](https://github.com/huggingface/diffusers/issues?q=is%3Aopen+is%3Aissue+label%3A%22New+scheduler%22) -Also, say ๐Ÿ‘‹ in our public Discord channel Join us on Discord. We discuss the hottest trends about diffusion models, help each other with contributions, personal projects or -just hang out โ˜•. +Also, say ๐Ÿ‘‹ in our public Discord channel Join us on Discord. We discuss the hottest trends about diffusion models, help each other with contributions, personal projects or just hang out โ˜•. ## Popular Tasks & Pipelines From 5c5bf6aba40a3ddbdbe78e5a37d1fe0eb774b421 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:32:06 +0300 Subject: [PATCH 3/9] Update memory.md --- docs/source/en/optimization/memory.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/docs/source/en/optimization/memory.md b/docs/source/en/optimization/memory.md index 281b65df8d8c..42a1bcea8fb5 100644 --- a/docs/source/en/optimization/memory.md +++ b/docs/source/en/optimization/memory.md @@ -194,9 +194,9 @@ unet_runs_per_experiment = 50 # load inputs def generate_inputs(): - sample = torch.randn(2, 4, 64, 64).half().cuda() - timestep = torch.rand(1).half().cuda() * 999 - encoder_hidden_states = torch.randn(2, 77, 768).half().cuda() + sample = torch.randn((2, 4, 64, 64), device="cuda", dtype=torch.float16) + timestep = torch.rand(1, device="cuda", dtype=torch.float16) * 999 + encoder_hidden_states = torch.randn((2, 77, 768), device="cuda", dtype=torch.float16) return sample, timestep, encoder_hidden_states From 6057f368b233a29a7a34acd6a19a769265da3da5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:34:04 +0300 Subject: [PATCH 4/9] Update basic_training.md --- docs/source/en/tutorials/basic_training.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/docs/source/en/tutorials/basic_training.md b/docs/source/en/tutorials/basic_training.md index 3b545cdf572e..ba7e0f01bf8d 100644 --- a/docs/source/en/tutorials/basic_training.md +++ b/docs/source/en/tutorials/basic_training.md @@ -321,13 +321,14 @@ Now you can wrap all these components together in a training loop with ๐Ÿค— Acce ... for step, batch in enumerate(train_dataloader): ... clean_images = batch["images"] ... # Sample noise to add to the images -... noise = torch.randn(clean_images.shape).to(clean_images.device) +... noise = torch.randn(clean_images.shape, device=clean_images.device) ... bs = clean_images.shape[0] ... # Sample a random timestep for each image ... timesteps = torch.randint( -... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device -... ).long() +... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device, +... dtype=torch.int64 +... ) ... # Add noise to the clean images according to the noise magnitude at each timestep ... # (this is the forward diffusion process) From fa2d8a4df5353167fecccb227a8b34862cba6993 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:35:30 +0300 Subject: [PATCH 5/9] Update write_own_pipeline.md --- docs/source/en/using-diffusers/write_own_pipeline.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/source/en/using-diffusers/write_own_pipeline.md b/docs/source/en/using-diffusers/write_own_pipeline.md index d3061e36fe63..4ca3fe33223b 100644 --- a/docs/source/en/using-diffusers/write_own_pipeline.md +++ b/docs/source/en/using-diffusers/write_own_pipeline.md @@ -71,7 +71,7 @@ tensor([980, 960, 940, 920, 900, 880, 860, 840, 820, 800, 780, 760, 740, 720, >>> import torch >>> sample_size = model.config.sample_size ->>> noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") +>>> noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") ``` 5. Now write a loop to iterate over the timesteps. At each timestep, the model does a [`UNet2DModel.forward`] pass and returns the noisy residual. The scheduler's [`~DDPMScheduler.step`] method takes the noisy residual, timestep, and input and it predicts the image at the previous timestep. This output becomes the next input to the model in the denoising loop, and it'll repeat until it reaches the end of the `timesteps` array. @@ -216,8 +216,8 @@ Next, generate some initial random noise as a starting point for the diffusion p >>> latents = torch.randn( ... (batch_size, unet.config.in_channels, height // 8, width // 8), ... generator=generator, +... device=torch_device, ... ) ->>> latents = latents.to(torch_device) ``` ### Denoise the image From 3b2b06ac57b6687019a6cce4cd67ca7a4eceba21 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:37:05 +0300 Subject: [PATCH 6/9] Update fp16.md --- docs/source/ko/optimization/fp16.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/docs/source/ko/optimization/fp16.md b/docs/source/ko/optimization/fp16.md index 30197305540c..0f2c487a75ce 100644 --- a/docs/source/ko/optimization/fp16.md +++ b/docs/source/ko/optimization/fp16.md @@ -273,9 +273,9 @@ unet_runs_per_experiment = 50 # ์ž…๋ ฅ ๋ถˆ๋Ÿฌ์˜ค๊ธฐ def generate_inputs(): - sample = torch.randn(2, 4, 64, 64).half().cuda() - timestep = torch.rand(1).half().cuda() * 999 - encoder_hidden_states = torch.randn(2, 77, 768).half().cuda() + sample = torch.randn((2, 4, 64, 64), device="cuda", dtype=torch.float16) + timestep = torch.rand(1, device="cuda", dtype=torch.float16) * 999 + encoder_hidden_states = torch.randn((2, 77, 768), device="cuda", dtype=torch.float16) return sample, timestep, encoder_hidden_states From 3dac1b5fb1385a6ec06112fef74cc383334ee1dd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:38:16 +0300 Subject: [PATCH 7/9] Update basic_training.md --- docs/source/ko/tutorials/basic_training.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/docs/source/ko/tutorials/basic_training.md b/docs/source/ko/tutorials/basic_training.md index a4e5e2a0c8bb..1cc82d2b8ce6 100644 --- a/docs/source/ko/tutorials/basic_training.md +++ b/docs/source/ko/tutorials/basic_training.md @@ -322,13 +322,14 @@ TensorBoard์— ๋กœ๊น…, ๊ทธ๋ž˜๋””์–ธํŠธ ๋ˆ„์  ๋ฐ ํ˜ผํ•ฉ ์ •๋ฐ€๋„ ํ•™์Šต์„ ์‰ฝ ... for step, batch in enumerate(train_dataloader): ... clean_images = batch["images"] ... # ์ด๋ฏธ์ง€์— ๋”ํ•  ๋…ธ์ด์ฆˆ๋ฅผ ์ƒ˜ํ”Œ๋งํ•ฉ๋‹ˆ๋‹ค. -... noise = torch.randn(clean_images.shape).to(clean_images.device) +... noise = torch.randn(clean_images.shape, device=clean_images.device) ... bs = clean_images.shape[0] ... # ๊ฐ ์ด๋ฏธ์ง€๋ฅผ ์œ„ํ•œ ๋žœ๋คํ•œ ํƒ€์ž„์Šคํ…(timestep)์„ ์ƒ˜ํ”Œ๋งํ•ฉ๋‹ˆ๋‹ค. ... timesteps = torch.randint( -... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device -... ).long() +... 0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device, +... dtype=torch.int64 +... ) ... # ๊ฐ ํƒ€์ž„์Šคํ…์˜ ๋…ธ์ด์ฆˆ ํฌ๊ธฐ์— ๋”ฐ๋ผ ๊นจ๋—ํ•œ ์ด๋ฏธ์ง€์— ๋…ธ์ด์ฆˆ๋ฅผ ์ถ”๊ฐ€ํ•ฉ๋‹ˆ๋‹ค. ... # (์ด๋Š” foward diffusion ๊ณผ์ •์ž…๋‹ˆ๋‹ค.) From 66fa836f9d0fbfa2c2e147f4e8ba8a6246b02f53 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:39:18 +0300 Subject: [PATCH 8/9] Update write_own_pipeline.md --- docs/source/ko/using-diffusers/write_own_pipeline.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/docs/source/ko/using-diffusers/write_own_pipeline.md b/docs/source/ko/using-diffusers/write_own_pipeline.md index a6469644566c..4f356e08629a 100644 --- a/docs/source/ko/using-diffusers/write_own_pipeline.md +++ b/docs/source/ko/using-diffusers/write_own_pipeline.md @@ -71,7 +71,7 @@ specific language governing permissions and limitations under the License. >>> import torch >>> sample_size = model.config.sample_size - >>> noise = torch.randn((1, 3, sample_size, sample_size)).to("cuda") + >>> noise = torch.randn((1, 3, sample_size, sample_size), device="cuda") ``` 5. ์ด์ œ timestep์„ ๋ฐ˜๋ณตํ•˜๋Š” ๋ฃจํ”„๋ฅผ ์ž‘์„ฑํ•ฉ๋‹ˆ๋‹ค. ๊ฐ timestep์—์„œ ๋ชจ๋ธ์€ [`UNet2DModel.forward`]๋ฅผ ํ†ตํ•ด noisy residual์„ ๋ฐ˜ํ™˜ํ•ฉ๋‹ˆ๋‹ค. ์Šค์ผ€์ค„๋Ÿฌ์˜ [`~DDPMScheduler.step`] ๋ฉ”์„œ๋“œ๋Š” noisy residual, timestep, ๊ทธ๋ฆฌ๊ณ  ์ž…๋ ฅ์„ ๋ฐ›์•„ ์ด์ „ timestep์—์„œ ์ด๋ฏธ์ง€๋ฅผ ์˜ˆ์ธกํ•ฉ๋‹ˆ๋‹ค. ์ด ์ถœ๋ ฅ์€ ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ ๋ฃจํ”„์˜ ๋ชจ๋ธ์— ๋Œ€ํ•œ ๋‹ค์Œ ์ž…๋ ฅ์ด ๋˜๋ฉฐ, `timesteps` ๋ฐฐ์—ด์˜ ๋์— ๋„๋‹ฌํ•  ๋•Œ๊นŒ์ง€ ๋ฐ˜๋ณต๋ฉ๋‹ˆ๋‹ค. @@ -212,6 +212,7 @@ Stable Diffusion ์€ text-to-image *latent diffusion* ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. latent di >>> latents = torch.randn( ... (batch_size, unet.in_channels, height // 8, width // 8), ... generator=generator, +... device=torch_device, ... ) >>> latents = latents.to(torch_device) ``` From 004b19cfcb2ddafc5111f2933f49210e33189061 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?M=2E=20Tolga=20Cang=C3=B6z?= <46008593+standardAI@users.noreply.github.com> Date: Sat, 18 Nov 2023 09:41:21 +0300 Subject: [PATCH 9/9] Update write_own_pipeline.md --- docs/source/ko/using-diffusers/write_own_pipeline.md | 1 - 1 file changed, 1 deletion(-) diff --git a/docs/source/ko/using-diffusers/write_own_pipeline.md b/docs/source/ko/using-diffusers/write_own_pipeline.md index 4f356e08629a..787c8113bf0d 100644 --- a/docs/source/ko/using-diffusers/write_own_pipeline.md +++ b/docs/source/ko/using-diffusers/write_own_pipeline.md @@ -214,7 +214,6 @@ Stable Diffusion ์€ text-to-image *latent diffusion* ๋ชจ๋ธ์ž…๋‹ˆ๋‹ค. latent di ... generator=generator, ... device=torch_device, ... ) ->>> latents = latents.to(torch_device) ``` ### ์ด๋ฏธ์ง€ ๋…ธ์ด์ฆˆ ์ œ๊ฑฐ