unet/config.json

{
  "_class_name": "UNetConditionalModel",
  "_diffusers_version": "0.0.4",
  "attention_resolutions": [
    4,
    2,
    1
  ],
  "attn_resolutions": null,
  "block_channels": [
    320,
    640,
    1280,
    1280
  ],
  "center_input_sample": false,
  "conv_resample": true,
  "down_blocks": [
    "UNetResCrossAttnDownBlock2D",
    "UNetResCrossAttnDownBlock2D",
    "UNetResCrossAttnDownBlock2D",
    "UNetResDownBlock2D"
  ],
  "downsample_padding": 1,
  "downscale_freq_shift": 0,
  "dropout": 0,
  "flip_sin_to_cos": true,
  "image_size": 32,
  "in_channels": 4,
  "mid_block_scale_factor": 1,
  "name_or_path": "fusing/latent-diffusion-text2im-large",
  "num_head_channels": 8,
	"num_attention_heads": 8,
  "num_res_blocks": 2,
  "out_channels": 4,
  "resnet_eps": 1e-05,
  "resnet_num_groups": 32,
  "up_blocks": [
    "UNetResUpBlock2D",
    "UNetResCrossAttnUpBlock2D",
    "UNetResCrossAttnUpBlock2D",
    "UNetResCrossAttnUpBlock2D"
  ]
}