bump diffusers version and update unet config

model_index.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "AudioLDMPipeline",
-  "_diffusers_version": "0.13.0.dev0",
+  "_diffusers_version": "0.14.0.dev0",
   "scheduler": [
     "diffusers",
     "DDIMScheduler"

scheduler/scheduler_config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "DDIMScheduler",
-  "_diffusers_version": "0.13.0.dev0",
+  "_diffusers_version": "0.14.0.dev0",
   "beta_end": 0.0195,
   "beta_schedule": "scaled_linear",
   "beta_start": 0.0015,

unet/config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "UNet2DConditionModel",
-  "_diffusers_version": "0.13.0.dev0",
+  "_diffusers_version": "0.14.0.dev0",
   "act_fn": "silu",
   "attention_head_dim": 8,
   "block_out_channels": [
@@ -13,7 +13,12 @@
   "class_embed_type": null,
   "conv_in_kernel": 3,
   "conv_out_kernel": 3,
-  "cross_attention_dim": true,
+  "cross_attention_dim": [
+    128,
+    256,
+    384,
+    640
+  ],
   "down_block_types": [
     "DownBlock2D",
     "CrossAttnDownBlock2D",
@@ -35,6 +40,7 @@
   "num_class_embeds": null,
   "only_cross_attention": false,
   "out_channels": 8,
+  "projection_class_embeddings_input_dim": null,
   "resnet_time_scale_shift": "default",
   "sample_size": 128,
   "time_cond_proj_dim": null,

vae/config.json

@@ -1,6 +1,6 @@
 {
   "_class_name": "AutoencoderKL",
-  "_diffusers_version": "0.13.0.dev0",
+  "_diffusers_version": "0.14.0.dev0",
   "act_fn": "silu",
   "block_out_channels": [
     128,