add: diffusers weights.

README.md

@@ -9,4 +9,21 @@ TAESDXL is useful for [real-time previewing](https://twitter.com/madebyollin/sta
 
 This repo contains `.safetensors` versions of the TAESDXL weights.
 
-For SD1.x / SD2.x, use [TAESD](https://huggingface.co/madebyollin/taesd/) instead (the SD and SDXL VAEs are [incompatible](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/discussions/6#64b8a9c13707b7d603c6ac16)).
+For SD1.x / SD2.x, use [TAESD](https://huggingface.co/madebyollin/taesd/) instead (the SD and SDXL VAEs are [incompatible](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/discussions/6#64b8a9c13707b7d603c6ac16)).
+
+## Using in 🧨 diffusers
+
+```python
+import torch
+from diffusers import DiffusionPipeline, AutoencoderTiny
+
+pipe = DiffusionPipeline.from_pretrained(
+    "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16
+)
+pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taesdxl", torch_dtype=torch.float16)
+pipe = pipe.to("cuda")
+
+prompt = "slice of delicious New York-style berry cheesecake"
+image = pipe(prompt, num_inference_steps=25).images[0]
+image.save("cheesecake_sdxl.png")
+```

config.json

@@ -0,0 +1,37 @@
+{
+  "_class_name": "AutoencoderTiny",
+  "_diffusers_version": "0.20.0.dev0",
+  "act_fn": "relu",
+  "decoder_block_out_channels": [
+    64,
+    64,
+    64,
+    64
+  ],
+  "encoder_block_out_channels": [
+    64,
+    64,
+    64,
+    64
+  ],
+  "force_upcast": false,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "latent_magnitude": 3,
+  "latent_shift": 0.5,
+  "num_decoder_blocks": [
+    3,
+    3,
+    3,
+    1
+  ],
+  "num_encoder_blocks": [
+    1,
+    3,
+    3,
+    3
+  ],
+  "out_channels": 3,
+  "scaling_factor": 1.0,
+  "upsampling_scaling_factor": 2
+}

diffusion_pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0af7e9e60c66873ac787d11f2955a1e1b5a6b7343954d92e74507b71f84f312c
+size 9832437

diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55ef8e24c0ef92d7f1d81b2cf66283d72396c6c05b6f073bec845147cd1247f3
+size 9793292