Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/nod-ai/sdxl-scripts

Last synced: 7 days ago
JSON representation

Host: GitHub
URL: https://github.com/nod-ai/sdxl-scripts
Owner: nod-ai
License: apache-2.0
Created: 2024-04-24T20:03:38.000Z (7 months ago)
Default Branch: main
Last Pushed: 2024-08-23T00:46:54.000Z (3 months ago)
Last Synced: 2024-08-23T19:02:05.028Z (3 months ago)
Language: MLIR
Size: 13.3 MB
Stars: 2
Watchers: 4
Forks: 5
Open Issues: 7
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

        # SDXL IRs and Scripts

## SDXL end-to-end benchmarking

1. Checkout and compile IREE with release build and `export PATH=/path/to/iree/build/release/tools:$PATH`

2. Compile the full SDXL model: `./compile-txt2img.sh gfx942` (where `gfx942` is the target for MI300X)

3. Run the benchmark: `./benchmark-txt2img.sh N /path/to/weights/irpa` (where `N` is the GPU index)

## Model IRs and weights

> [!CAUTION]

> IRs in the following table might be stale. Use the ones in the

> `base_ir/` directory instead.

> [!NOTE]

> SDXL-turbo is only different from SDXL in its usage and training/weights.

> The model architecture (and therefore the weights-stripped MLIR) are equivalent.

Variant | Submodel | MLIR (No Weights) (Config A) | safetensors | Splat IRPA | MLIR (No Weights) (Config B)

:-----: | :---------: | :-------------: | :-------------: | :------: | :--------:

| SDXL1.0 1024x1024 (f16, BS1, len64) |

|  | UNet + attn | [Torch](https://sharkpublic.blob.core.windows.net/sharkpublic/ian/stable_diffusion_xl_base_1_0_64_1024x1024_fp16_unet_torchir_nithin.mlir) - [Linalg](https://sharkpublic.blob.core.windows.net/sharkpublic/ian/stable_diffusion_xl_base_1_0_64_1024x1024_fp16_unet_linalg_nithin.mlir)| - | - | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_b/fp16/stable_diffusion_xl_base_1_0_64_1024x1024_fp16_unet.mlir)

|  | UNet + PNDMScheduler | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/stable_diffusion_xl_base_1_0_scheduler_rocm.mlir)

|  | Clip1 | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp16/stable_diffusion_xl_base_1_0_64_fp16_clip_1.mlir) | - | - |

|  | Clip2 | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp16/stable_diffusion_xl_base_1_0_64_fp16_clip_2.mlir) | - | - |

|  | VAE decode + attn | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp16/stable_diffusion_xl_base_1_0_1024x1024_fp16_vae_decode.mlir) | - | = | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_b/fp16/stable_diffusion_xl_base_1_0_1024x1024_fp16_vae_decode.mlir)

|  | VAE encode + attn | [GCloud][sdxl-1-1024x1024-f16-stripped-weight-vae-encode] | Same as decode | - | -

| SDXL1.0 1024x1024 (f32, BS1, len64) |

|  | UNet + attn | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_64_1024x1024_fp32_unet.mlir) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_unet.safetensors) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_unet.irpa) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_b/fp32/stable_diffusion_xl_base_1_0_64_1024x1024_fp32_unet.mlir)

|  | Clip1 | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_64_fp32_clip_1.mlir) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_clip_1.safetensors) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_clip_1.irpa) | -

|  | Clip2 | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_64_fp32_clip_2.mlir) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_clip_2.safetensors) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_clip_2.irpa) | -

|  | VAE decode + attn | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_1024x1024_fp32_vae_decode.mlir) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_vae_decode.safetensors) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_a/fp32/stable_diffusion_xl_base_1_0_fp32_vae_decode.irpa) | [Azure](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/cfg_b/fp32/stable_diffusion_xl_base_1_0_1024x1024_fp32_vae_decode.mlir)

| SDXL compiled pipeline IRPAs (f16) |

|  | UNet  |  [scheduled_unet_f16.irpa](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/SDXL_weights_fp16/scheduled_unet_fp16.irpa)

|  | Prompt Encoder (CLIP1 + CLIP2) | [prompt_encoder_f16.irpa](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/SDXL_weights_fp16/prompt_encoder_fp16.irpa)

|  | VAE |  [vae_decode_f16.irpa](https://sharkpublic.blob.core.windows.net/sharkpublic/SDXL/SDXL_weights_fp16/vae_decode_fp16.irpa)