Spaces:
Runtime error
Runtime error
| model: | |
| base_learning_rate: 2.0e-06 | |
| target: ldm.models.diffusion.ddpm.LatentDiffusion | |
| params: | |
| linear_start: 0.0015 | |
| linear_end: 0.0195 | |
| num_timesteps_cond: 1 | |
| log_every_t: 100 | |
| timesteps: 1000 | |
| first_stage_key: "target_image" | |
| cond_stage_key: "source_image" | |
| cond_stage_trainable: False | |
| cond_stage_instruction_key: "text" | |
| cond_stage_instruction_embedder_trainable: True | |
| conditioning_key: "hybrid" | |
| image_size: 32 | |
| channels: 4 | |
| monitor: val/loss_simple_ema | |
| unet_config: | |
| target: ldm.modules.diffusionmodules.openaimodel.UNetModel | |
| params: | |
| image_size: 32 | |
| in_channels: 8 # 4 (noisy image features) + 4 (source image features) | |
| out_channels: 4 | |
| model_channels: 128 | |
| attention_resolutions: [8,4,2] | |
| num_res_blocks: 2 | |
| channel_mult: [1,2,3,4] | |
| num_heads: 8 | |
| resblock_updown: True | |
| ###### Instruction embedding cross attention ###### | |
| use_spatial_transformer: true | |
| transformer_depth: 1 | |
| context_dim: 512 | |
| ################################################### | |
| first_stage_config: | |
| target: ldm.models.autoencoder.VQModel | |
| params: | |
| ckpt_path: models/gqa_inpaint/first_stage/vq-f8-cb16384-openimages.ckpt | |
| monitor: "val/rec_loss" | |
| embed_dim: 4 | |
| n_embed: 16384 | |
| lossconfig: | |
| target: torch.nn.Identity | |
| ddconfig: | |
| double_z: false | |
| z_channels: 4 | |
| resolution: 256 | |
| in_channels: 3 | |
| out_ch: 3 | |
| ch: 128 | |
| ch_mult: [1,2,2,4] | |
| num_res_blocks: 2 | |
| attn_resolutions: [32] | |
| dropout: 0.0 | |
| cond_stage_config: __is_first_stage__ | |
| cond_stage_instruction_embedder_config: | |
| target: ldm.modules.encoders.modules.BERTEmbedder | |
| params: | |
| n_embed: 512 | |
| n_layer: 16 | |
| data: | |
| target: main.DataModuleFromConfig | |
| params: | |
| batch_size: 8 | |
| num_workers: 4 | |
| train: | |
| target: dataset.gqa_inpaint.GQAInpaintTrain | |
| params: | |
| images_root: "data/gqa-inpaint/images" | |
| images_inpainted_root: "data/gqa-inpaint/images_inpainted" | |
| masks_root: "data/gqa-inpaint/masks" | |
| scene_json_path: "data/gqa-inpaint/train_scenes.json" | |
| max_relations: 1 | |
| simplify_augment: True | |
| instruction_type: "remove" | |
| size: 256 | |
| irrelevant_text_prob: 0.2 | |
| validation: | |
| target: dataset.gqa_inpaint.GQAInpaintTest | |
| params: | |
| images_root: "data/gqa-inpaint/images" | |
| images_inpainted_root: "data/gqa-inpaint/images_inpainted" | |
| masks_root: "data/gqa-inpaint/masks" | |
| scene_json_path: "data/gqa-inpaint/test_scenes.json" | |
| max_relations: 1 | |
| simplify_augment: True | |
| instruction_type: "remove" | |
| size: 256 | |
| test: | |
| target: dataset.gqa_inpaint.GQAInpaintTest | |
| params: | |
| images_root: "data/gqa-inpaint/images" | |
| images_inpainted_root: "data/gqa-inpaint/images_inpainted" | |
| masks_root: "data/gqa-inpaint/masks" | |
| scene_json_path: "data/gqa-inpaint/test_scenes.json" | |
| test_instructions_path: "data/gqa-inpaint/test_instructions.json" | |
| max_relations: 1 | |
| simplify_augment: True | |
| instruction_type: "remove" | |
| size: 256 | |
| lightning: | |
| callbacks: | |
| image_logger: | |
| target: main.ImageLogger | |
| params: | |
| batch_frequency: 5000 | |
| max_images: 8 | |
| increase_log_steps: True | |
| trainer: | |
| benchmark: True |