| { | |
| "architectures": [ | |
| "QueryTransformerEmbedding" | |
| ], | |
| "attention_probs_dropout_prob": 0.0, | |
| "hidden_act": "gelu_fast", | |
| "hidden_dropout_prob": 0.0, | |
| "hidden_size": 768, | |
| "image_size": [ | |
| 72, | |
| 128 | |
| ], | |
| "initializer_range": 0.02, | |
| "intermediate_size": 3072, | |
| "layer_norm_eps": 1e-06, | |
| "model_type": "vivit", | |
| "num_attention_heads": 8, | |
| "num_channels": 8, | |
| "num_frames": 24, | |
| "num_hidden_layers": 6, | |
| "qkv_bias": true, | |
| "query_dims": [ | |
| 160, | |
| 320, | |
| 640 | |
| ], | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.47.1", | |
| "tubelet_size": [ | |
| 2, | |
| 9, | |
| 16 | |
| ], | |
| "unet_attention_names": { | |
| "down_blocks_0_attentions_0_temporal_transformer_blocks_0_attn1": 320, | |
| "down_blocks_0_attentions_0_transformer_blocks_0_attn1": 320, | |
| "down_blocks_0_attentions_1_temporal_transformer_blocks_0_attn1": 320, | |
| "down_blocks_0_attentions_1_transformer_blocks_0_attn1": 320, | |
| "down_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 640, | |
| "down_blocks_1_attentions_0_transformer_blocks_0_attn1": 640, | |
| "down_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 640, | |
| "down_blocks_1_attentions_1_transformer_blocks_0_attn1": 640, | |
| "down_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 1280, | |
| "down_blocks_2_attentions_0_transformer_blocks_0_attn1": 1280, | |
| "down_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 1280, | |
| "down_blocks_2_attentions_1_transformer_blocks_0_attn1": 1280, | |
| "mid_block_attentions_0_temporal_transformer_blocks_0_attn1": 1280, | |
| "mid_block_attentions_0_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_0_temporal_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_0_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_1_temporal_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_1_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_2_temporal_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_1_attentions_2_transformer_blocks_0_attn1": 1280, | |
| "up_blocks_2_attentions_0_temporal_transformer_blocks_0_attn1": 640, | |
| "up_blocks_2_attentions_0_transformer_blocks_0_attn1": 640, | |
| "up_blocks_2_attentions_1_temporal_transformer_blocks_0_attn1": 640, | |
| "up_blocks_2_attentions_1_transformer_blocks_0_attn1": 640, | |
| "up_blocks_2_attentions_2_temporal_transformer_blocks_0_attn1": 640, | |
| "up_blocks_2_attentions_2_transformer_blocks_0_attn1": 640, | |
| "up_blocks_3_attentions_0_temporal_transformer_blocks_0_attn1": 320, | |
| "up_blocks_3_attentions_0_transformer_blocks_0_attn1": 320, | |
| "up_blocks_3_attentions_1_temporal_transformer_blocks_0_attn1": 320, | |
| "up_blocks_3_attentions_1_transformer_blocks_0_attn1": 320, | |
| "up_blocks_3_attentions_2_temporal_transformer_blocks_0_attn1": 320, | |
| "up_blocks_3_attentions_2_transformer_blocks_0_attn1": 320 | |
| }, | |
| "zero_init_theta": false | |
| } | |