Update unet/config.json

unet/config.json

@@ -1,6 +1,6 @@
 {
  "_class_name": "UNet2DConditionModel",
- "_diffusers_version": "0.19.0.dev0",
+ "_diffusers_version": "0.19.0",
  "act_fn": "silu",
  "addition_embed_type": "text_time",
  "addition_embed_type_num_heads": 64,
@@ -18,13 +18,14 @@
  "center_input_sample": false,
  "class_embed_type": null,
  "class_embeddings_concat": false,
- "conv_in_kernel": 3,
- "conv_out_kernel": 3,
+ "conv_in_kernel": 4,
+ "conv_out_kernel": 4,
  "cross_attention_dim": 2048,
  "cross_attention_norm": null,
  "down_block_types": [
  "DownBlock2D",
  "CrossAttnDownBlock2D",
+ "CrossAttnDownBlock2D",
  "CrossAttnDownBlock2D"
  ],
  "downsample_padding": 1,
@@ -37,7 +38,7 @@
  "layers_per_block": 2,
  "mid_block_only_cross_attention": null,
  "mid_block_scale_factor": 1,
- "mid_block_type": "UNetMidBlock2D",
+ "mid_block_type": "UNetMidBlock2DCrossAttn",
  "norm_eps": 1e-05,
  "norm_num_groups": 32,
  "num_attention_heads": null,
@@ -55,14 +56,9 @@
  "time_embedding_type": "positional",
  "timestep_post_act": null,
  "transformer_layers_per_block": [
- [1],
- [2,2],
- [4,4]
- ],
- "reverse_transformer_layers_per_block": [
- [4,4,10],
- [2,1,1],
- 1
+ 1,
+ 2,
+ 10
  ],
  "up_block_types": [
  "CrossAttnUpBlock2D",