add new model, train unet motion and ipadapter, fix referencenet, t2i

motion/musev_referencenet_pose/ip_adapter_image_proj.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f82b521bcf7831558ce2cc97721e84dbf36b793c2ec9fae16d63a8b2f0f7508
+size 12603123

motion/musev_referencenet_pose/unet/config.json

@@ -0,0 +1,71 @@
+{
+  "_class_name": "UNet3DConditionModel",
+  "_diffusers_version": "0.24.0.dev0",
+  "_name_or_path": "/cfs-datasets/public_models/fantasticmix_v10",
+  "act_fn": "silu",
+  "attention_head_dim": 8,
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "center_input_sample": false,
+  "class_embed_type": null,
+  "class_embeddings_concat": false,
+  "conv_in_kernel": 3,
+  "conv_out_kernel": 3,
+  "cross_attention_dim": 768,
+  "down_block_types": [
+    "CrossAttnDownBlock3D",
+    "CrossAttnDownBlock3D",
+    "CrossAttnDownBlock3D",
+    "DownBlock3D"
+  ],
+  "downsample_padding": 1,
+  "dual_cross_attention": false,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 4,
+  "ip_adapter_cross_attn": true,
+  "keep_vision_condtion": true,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock3DCrossAttn",
+  "need_adain_temporal_cond": true,
+  "need_refer_emb": true,
+  "need_spatial_position_emb": false,
+  "need_t2i_facein": false,
+  "need_t2i_ip_adapter": true,
+  "need_t2i_ip_adapter_face": false,
+  "need_transformer_in": false,
+  "need_vis_cond_mask": false,
+  "need_zero_vis_cond_temb": true,
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "norm_spatial_length": true,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "out_channels": 4,
+  "projection_class_embeddings_input_dim": null,
+  "resnet_2d_skip_time_act": true,
+  "resnet_time_scale_shift": "default",
+  "sample_size": 64,
+  "spatial_max_length": 2048,
+  "t2i_crossattn_ip_adapter_attn_processor": "T2IReferencenetIPAdapterXFormersAttnProcessor",
+  "t2i_ip_adapter_attn_processor": "NonParamReferenceIPXFormersAttnProcessor",
+  "temporal_conv_block": "TemporalConvLayer",
+  "temporal_transformer": "TransformerTemporalModel",
+  "time_cond_proj_dim": null,
+  "time_embedding_type": "positional",
+  "timestep_post_act": null,
+  "up_block_types": [
+    "UpBlock3D",
+    "CrossAttnUpBlock3D",
+    "CrossAttnUpBlock3D",
+    "CrossAttnUpBlock3D"
+  ],
+  "upcast_attention": false,
+  "use_anivv1_cfg": true,
+  "use_linear_projection": false
+}

motion/musev_referencenet_pose/unet/diffusion_pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69dd5cd26d9ff46b6c5e8879ed9b82aa7606810755cb2a9e5b3ed785242d0ebc
+size 5930137448