Skip to content

Releases: mlcommons/mobile_open

v4.1: ONNX models for Stable Diffusion

16 Oct 09:13
2984685
Compare
Choose a tag to compare

The Stable Diffusion v1.5 models are converted to ONNX. Time Embeddings are precomputed, where the model for Unet takes 1x1280 precomputed Time Embeddings.

# SHA256 checksum
642cbf6faf373b2bc1a180c1ba7868cbd675db82da7dd8108ceaa586f8b67a42  text_encoder.onnx
645d40e7e7eb9e8a669390d8bb6e14377679b7376da55fb3b4c1e6e551e4de15  unet.onnx
988c3a4cadaef719f7f4c85bd97eafc37980c16eeddce86dc9664f06fc19c7ee  vae_decoder.onnx

v4.1: TFLite models for Stable Diffusion

07 Jan 12:48
2984685
Compare
Choose a tag to compare
# SHA 256
f5386ed5848cd50cd885c3a247369395be579ede5bba2660181ec13c3060c112  TEXT_ENCODER_sdv15_fixed_batch_fp32.tflite
e439951df040a2251ce0c8e30847a720d6172c623eb2beb757a8bb994f53b2f1  TIME_EMBED_HEAD_sdv15_time_emb_head_LinSiluLin_fp32.tflite
3e589a586029f015c54eb7f805146bbad1651e460ccae1d49e2756e96e2c32d2  UNET_p1_sdv15_diffusion_model_first_fp32.tflite
0b366ef4fd18f730983696486afdbc73235e7f5ed700db6ca55e1cd417b9ea30  UNET_p2_sdv15_diffusion_model_second_fp32.tflite
21e1df52ee58abd910370f94579cd9dcbc9c34fe59e0747b7308481a5220934f  VAE_DECODER_sdv15_fixed_batch_fp32.tflite
181637b7f8605a9a430d1537807effe4e407bb2daa7d223254fbd74ac14d3774  sd_decoder_qint8.tflite
92eb8431100c36b7a19fb4513407db5f822a1818fd4bb473461a03869e29461c  sd_diffusion_full_model_dynamic_qint8.tflite
1bfb962195c569e64c3506a2c096f18fa925bbb883bd2e4eb88f9a4950685851  sd_text_encoder_qint8.tflite
4eebd2fe09a0c67157f996037aaa1ef2d486f30794f012ebc5019c26deea8c84  timestep_embeddings_data.bin

MobileNetV4 - Saved Model and FP23 TFLite

07 Feb 23:38
Compare
Choose a tag to compare

Uploading the MobileNetV4-Conv-Large Saved Model and FP23 TFLite.