| { | |
| "_name_or_path": "solintellegence/SolPix", | |
| "license": "apache-2.0", | |
| "release": "v1.0.0", | |
| "architectures": [ | |
| "SolPixTransformer2D" | |
| ], | |
| "model_type": "solpix", | |
| "model_config": { | |
| "latent_channels": 32, | |
| "text_dim": 768, | |
| "max_text_tokens": 96, | |
| "width": 512, | |
| "depth": 15, | |
| "heads": 8, | |
| "mlp_hidden": 1152, | |
| "encoder_depth": 7, | |
| "latent_downsample_factor": 32, | |
| "image_channels": 3, | |
| "repa_layer": 7 | |
| }, | |
| "transformer_class": "solpix.SolPixTransformer2D", | |
| "prediction_type": "flow_velocity", | |
| "text_encoder": { | |
| "class": "transformers.T5EncoderModel", | |
| "model_id": "google/flan-t5-base", | |
| "hidden_size": 768, | |
| "max_tokens": 96 | |
| }, | |
| "autoencoder": { | |
| "class": "solpix.AutoencoderDCSol", | |
| "implementation": "diffusers.AutoencoderDC", | |
| "model_id": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers", | |
| "revision": "df0d9d634aea77793c1fb685d4b9db092c99e686", | |
| "latent_channels": 32, | |
| "downsample_factor": 32 | |
| }, | |
| "checkpoint": { | |
| "filename": "step_00210000.pt", | |
| "optimizer_step": 210000, | |
| "ema_weights_key": "ema" | |
| } | |
| } | |