Update transformer_I2V/config.json

transformer_I2V/config.json

@@ -1,32 +1,32 @@
-{
-  "_class_name": "CogVideoXTransformer3DModel",
-  "_diffusers_version": "0.31.0",
-  "activation_fn": "gelu-approximate",
-  "attention_bias": true,
-  "attention_head_dim": 64,
-  "dropout": 0.0,
-  "flip_sin_to_cos": true,
-  "freq_shift": 0,
-  "in_channels": 32,
-  "max_text_seq_length": 226,
-  "norm_elementwise_affine": true,
-  "norm_eps": 1e-05,
-  "num_attention_heads": 48,
-  "num_layers": 42,
-  "ofs_embed_dim": 512,
-  "out_channels": 16,
-  "patch_bias": false,
-  "patch_size": 2,
-  "patch_size_t": 2,
-  "sample_frames": 81,
-  "sample_height": 96,
-  "sample_width": 170,
-  "spatial_interpolation_scale": 1.875,
-  "temporal_compression_ratio": 4,
-  "temporal_interpolation_scale": 1.0,
-  "text_embed_dim": 4096,
-  "time_embed_dim": 512,
-  "timestep_activation_fn": "silu",
-  "use_learned_positional_embeddings": false,
-  "use_rotary_positional_embeddings": true
-}
+{
+  "_class_name": "CogVideoXTransformer3DModel",
+  "_diffusers_version": "0.31.0",
+  "activation_fn": "gelu-approximate",
+  "attention_bias": true,
+  "attention_head_dim": 64,
+  "dropout": 0.0,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "in_channels": 32,
+  "max_text_seq_length": 226,
+  "norm_elementwise_affine": true,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 48,
+  "num_layers": 42,
+  "ofs_embed_dim": 512,
+  "out_channels": 16,
+  "patch_bias": false,
+  "patch_size": 2,
+  "patch_size_t": 2,
+  "sample_frames": 81,
+  "sample_height": 300,
+  "sample_width": 300,
+  "spatial_interpolation_scale": 1.875,
+  "temporal_compression_ratio": 4,
+  "temporal_interpolation_scale": 1.0,
+  "text_embed_dim": 4096,
+  "time_embed_dim": 512,
+  "timestep_activation_fn": "silu",
+  "use_learned_positional_embeddings": false,
+  "use_rotary_positional_embeddings": true
+}