Push Keras model using huggingface_hub.

README.md

@@ -22,24 +22,20 @@ The following hyperparameters were used during training:
 
 | Hyperparameters | Value |
 | :-- | :-- |
-| inner_optimizer.class_name | Custom>RMSprop |
-| inner_optimizer.config.name | RMSprop |
-| inner_optimizer.config.weight_decay | None |
-| inner_optimizer.config.clipnorm | None |
-| inner_optimizer.config.global_clipnorm | None |
-| inner_optimizer.config.clipvalue | None |
-| inner_optimizer.config.use_ema | False |
-| inner_optimizer.config.ema_momentum | 0.99 |
-| inner_optimizer.config.ema_overwrite_frequency | 100 |
-| inner_optimizer.config.jit_compile | True |
-| inner_optimizer.config.is_legacy_optimizer | False |
-| inner_optimizer.config.learning_rate | 0.0010000000474974513 |
-| inner_optimizer.config.rho | 0.9 |
-| inner_optimizer.config.momentum | 0.0 |
-| inner_optimizer.config.epsilon | 1e-07 |
-| inner_optimizer.config.centered | False |
-| dynamic | True |
-| initial_scale | 32768.0 |
-| dynamic_growth_steps | 2000 |
-| training_precision | mixed_float16 |
+| name | RMSprop |
+| weight_decay | None |
+| clipnorm | None |
+| global_clipnorm | None |
+| clipvalue | None |
+| use_ema | False |
+| ema_momentum | 0.99 |
+| ema_overwrite_frequency | 100 |
+| jit_compile | True |
+| is_legacy_optimizer | False |
+| learning_rate | 0.0010000000474974513 |
+| rho | 0.9 |
+| momentum | 0.0 |
+| epsilon | 1e-07 |
+| centered | False |
+| training_precision | float32 |
 

config.json

@@ -1 +1 @@
-{"name": "diffusion_model_v2_1", "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": [null, 320], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_6"}, "name": "input_6", "inbound_nodes": []}, {"class_name": "Dense", "config": {"name": "dense_354", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "units": 1280, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_354", "inbound_nodes": [[["input_6", 0, 0, {}]]]}, {"class_name": "InputLayer", "config": {"batch_input_shape": [null, 64, 64, 4], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_7"}, "name": "input_7", "inbound_nodes": []}, {"class_name": "Activation", "config": {"name": "activation_69", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "activation": "swish"}, "name": "activation_69", "inbound_nodes": [[["dense_354", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_82", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "kernel_size": 3, "padding": 1, "filters": 320}, "name": "padded_conv2d_82", "inbound_nodes": [[["input_7", 0, 0, {}]]]}, {"class_name": "Dense", "config": {"name": "dense_355", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "units": 1280, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_355", "inbound_nodes": [[["activation_69", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_22", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 320}, "name": "res_block_22", "inbound_nodes": [[["padded_conv2d_82", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "InputLayer", "config": {"batch_input_shape": [null, 77, 1024], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_5"}, "name": "input_5", "inbound_nodes": []}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_16", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_16", "inbound_nodes": [[["res_block_22", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_23", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 320}, "name": "res_block_23", "inbound_nodes": [[["spatial_transformer_16", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_17", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_17", "inbound_nodes": [[["res_block_23", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_87", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "strides": 2, "padding": 1, "filters": 320, "kernel_size": 3}, "name": "padded_conv2d_87", "inbound_nodes": [[["spatial_transformer_17", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_24", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 640}, "name": "res_block_24", "inbound_nodes": [[["padded_conv2d_87", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_18", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_18", "inbound_nodes": [[["res_block_24", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_25", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 640}, "name": "res_block_25", "inbound_nodes": [[["spatial_transformer_18", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_19", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_19", "inbound_nodes": [[["res_block_25", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_92", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "strides": 2, "padding": 1, "filters": 640, "kernel_size": 3}, "name": "padded_conv2d_92", "inbound_nodes": [[["spatial_transformer_19", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_26", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_26", "inbound_nodes": [[["padded_conv2d_92", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_20", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_20", "inbound_nodes": [[["res_block_26", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_27", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_27", "inbound_nodes": [[["spatial_transformer_20", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_21", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_21", "inbound_nodes": [[["res_block_27", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_97", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "strides": 2, "padding": 1, "filters": 1280, "kernel_size": 3}, "name": "padded_conv2d_97", "inbound_nodes": [[["spatial_transformer_21", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_28", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_28", "inbound_nodes": [[["padded_conv2d_97", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_29", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_29", "inbound_nodes": [[["res_block_28", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_30", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_30", "inbound_nodes": [[["res_block_29", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_22", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_22", "inbound_nodes": [[["res_block_30", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_31", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_31", "inbound_nodes": [[["spatial_transformer_22", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_12", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_12", "inbound_nodes": [[["res_block_31", 0, 0, {}], ["res_block_29", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_32", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_32", "inbound_nodes": [[["concatenate_12", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_13", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_13", "inbound_nodes": [[["res_block_32", 0, 0, {}], ["res_block_28", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_33", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_33", "inbound_nodes": [[["concatenate_13", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_14", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_14", "inbound_nodes": [[["res_block_33", 0, 0, {}], ["padded_conv2d_97", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_34", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_34", "inbound_nodes": [[["concatenate_14", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample_3", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "channels": 1280}, "name": "upsample_3", "inbound_nodes": [[["res_block_34", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_15", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_15", "inbound_nodes": [[["upsample_3", 0, 0, {}], ["spatial_transformer_21", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_35", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_35", "inbound_nodes": [[["concatenate_15", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_23", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_23", "inbound_nodes": [[["res_block_35", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_16", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_16", "inbound_nodes": [[["spatial_transformer_23", 0, 0, {}], ["spatial_transformer_20", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_36", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_36", "inbound_nodes": [[["concatenate_16", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_24", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_24", "inbound_nodes": [[["res_block_36", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_17", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_17", "inbound_nodes": [[["spatial_transformer_24", 0, 0, {}], ["padded_conv2d_92", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_37", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 1280}, "name": "res_block_37", "inbound_nodes": [[["concatenate_17", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_25", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_25", "inbound_nodes": [[["res_block_37", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample_4", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "channels": 1280}, "name": "upsample_4", "inbound_nodes": [[["spatial_transformer_25", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_18", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_18", "inbound_nodes": [[["upsample_4", 0, 0, {}], ["spatial_transformer_19", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_38", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 640}, "name": "res_block_38", "inbound_nodes": [[["concatenate_18", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_26", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_26", "inbound_nodes": [[["res_block_38", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_19", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_19", "inbound_nodes": [[["spatial_transformer_26", 0, 0, {}], ["spatial_transformer_18", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_39", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 640}, "name": "res_block_39", "inbound_nodes": [[["concatenate_19", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_27", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_27", "inbound_nodes": [[["res_block_39", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_20", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_20", "inbound_nodes": [[["spatial_transformer_27", 0, 0, {}], ["padded_conv2d_87", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_40", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 640}, "name": "res_block_40", "inbound_nodes": [[["concatenate_20", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_28", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_28", "inbound_nodes": [[["res_block_40", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample_5", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "channels": 640}, "name": "upsample_5", "inbound_nodes": [[["spatial_transformer_28", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_21", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_21", "inbound_nodes": [[["upsample_5", 0, 0, {}], ["spatial_transformer_17", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_41", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 320}, "name": "res_block_41", "inbound_nodes": [[["concatenate_21", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_29", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_29", "inbound_nodes": [[["res_block_41", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_22", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_22", "inbound_nodes": [[["spatial_transformer_29", 0, 0, {}], ["spatial_transformer_16", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_42", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 320}, "name": "res_block_42", "inbound_nodes": [[["concatenate_22", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_30", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_30", "inbound_nodes": [[["res_block_42", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_23", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "axis": -1}, "name": "concatenate_23", "inbound_nodes": [[["spatial_transformer_30", 0, 0, {}], ["padded_conv2d_82", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_43", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "output_dim": 320}, "name": "res_block_43", "inbound_nodes": [[["concatenate_23", 0, 0, {}], ["dense_355", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_31", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_31", "inbound_nodes": [[["res_block_43", 0, 0, {}], ["input_5", 0, 0, {}]]]}, {"class_name": "GroupNormalization", "config": {"name": "group_normalization_143", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "groups": 32, "axis": -1, "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}}, "gamma_initializer": {"class_name": "Ones", "config": {}}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "name": "group_normalization_143", "inbound_nodes": [[["spatial_transformer_31", 0, 0, {}]]]}, {"class_name": "Activation", "config": {"name": "activation_136", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "activation": "swish"}, "name": "activation_136", "inbound_nodes": [[["group_normalization_143", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_133", "trainable": true, "dtype": {"class_name": "Policy", "config": {"name": "mixed_float16"}, "shared_object_id": 1}, "kernel_size": 3, "padding": 1, "filters": 4}, "name": "padded_conv2d_133", "inbound_nodes": [[["activation_136", 0, 0, {}]]]}], "input_layers": [["input_7", 0, 0], ["input_6", 0, 0], ["input_5", 0, 0]], "output_layers": [["padded_conv2d_133", 0, 0]]}
+{"name": "diffusion_model_v2", "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": [null, 320], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_2"}, "name": "input_2", "inbound_nodes": []}, {"class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1280, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense", "inbound_nodes": [[["input_2", 0, 0, {}]]]}, {"class_name": "InputLayer", "config": {"batch_input_shape": [null, 64, 64, 4], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_3"}, "name": "input_3", "inbound_nodes": []}, {"class_name": "Activation", "config": {"name": "activation", "trainable": true, "dtype": "float32", "activation": "swish"}, "name": "activation", "inbound_nodes": [[["dense", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d", "trainable": true, "dtype": "float32", "kernel_size": 3, "padding": 1, "filters": 320}, "name": "padded_conv2d", "inbound_nodes": [[["input_3", 0, 0, {}]]]}, {"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "dtype": "float32", "units": 1280, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense_1", "inbound_nodes": [[["activation", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block", "trainable": true, "dtype": "float32", "output_dim": 320}, "name": "res_block", "inbound_nodes": [[["padded_conv2d", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "InputLayer", "config": {"batch_input_shape": [null, 77, 1024], "dtype": "float32", "sparse": false, "ragged": false, "name": "input_1"}, "name": "input_1", "inbound_nodes": []}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer", "inbound_nodes": [[["res_block", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_1", "trainable": true, "dtype": "float32", "output_dim": 320}, "name": "res_block_1", "inbound_nodes": [[["spatial_transformer", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_1", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_1", "inbound_nodes": [[["res_block_1", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_5", "trainable": true, "dtype": "float32", "strides": 2, "padding": 1, "filters": 320, "kernel_size": 3}, "name": "padded_conv2d_5", "inbound_nodes": [[["spatial_transformer_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_2", "trainable": true, "dtype": "float32", "output_dim": 640}, "name": "res_block_2", "inbound_nodes": [[["padded_conv2d_5", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_2", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_2", "inbound_nodes": [[["res_block_2", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_3", "trainable": true, "dtype": "float32", "output_dim": 640}, "name": "res_block_3", "inbound_nodes": [[["spatial_transformer_2", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_3", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_3", "inbound_nodes": [[["res_block_3", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_10", "trainable": true, "dtype": "float32", "strides": 2, "padding": 1, "filters": 640, "kernel_size": 3}, "name": "padded_conv2d_10", "inbound_nodes": [[["spatial_transformer_3", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_4", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_4", "inbound_nodes": [[["padded_conv2d_10", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_4", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_4", "inbound_nodes": [[["res_block_4", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_5", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_5", "inbound_nodes": [[["spatial_transformer_4", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_5", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_5", "inbound_nodes": [[["res_block_5", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_15", "trainable": true, "dtype": "float32", "strides": 2, "padding": 1, "filters": 1280, "kernel_size": 3}, "name": "padded_conv2d_15", "inbound_nodes": [[["spatial_transformer_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_6", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_6", "inbound_nodes": [[["padded_conv2d_15", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_7", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_7", "inbound_nodes": [[["res_block_6", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_8", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_8", "inbound_nodes": [[["res_block_7", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_6", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_6", "inbound_nodes": [[["res_block_8", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_9", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_9", "inbound_nodes": [[["spatial_transformer_6", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate", "inbound_nodes": [[["res_block_9", 0, 0, {}], ["res_block_7", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_10", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_10", "inbound_nodes": [[["concatenate", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_1", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_1", "inbound_nodes": [[["res_block_10", 0, 0, {}], ["res_block_6", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_11", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_11", "inbound_nodes": [[["concatenate_1", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_2", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_2", "inbound_nodes": [[["res_block_11", 0, 0, {}], ["padded_conv2d_15", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_12", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_12", "inbound_nodes": [[["concatenate_2", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample", "trainable": true, "dtype": "float32", "channels": 1280}, "name": "upsample", "inbound_nodes": [[["res_block_12", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_3", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_3", "inbound_nodes": [[["upsample", 0, 0, {}], ["spatial_transformer_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_13", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_13", "inbound_nodes": [[["concatenate_3", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_7", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_7", "inbound_nodes": [[["res_block_13", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_4", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_4", "inbound_nodes": [[["spatial_transformer_7", 0, 0, {}], ["spatial_transformer_4", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_14", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_14", "inbound_nodes": [[["concatenate_4", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_8", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_8", "inbound_nodes": [[["res_block_14", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_5", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_5", "inbound_nodes": [[["spatial_transformer_8", 0, 0, {}], ["padded_conv2d_10", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_15", "trainable": true, "dtype": "float32", "output_dim": 1280}, "name": "res_block_15", "inbound_nodes": [[["concatenate_5", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_9", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 20, "head_size": 64}, "name": "spatial_transformer_9", "inbound_nodes": [[["res_block_15", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample_1", "trainable": true, "dtype": "float32", "channels": 1280}, "name": "upsample_1", "inbound_nodes": [[["spatial_transformer_9", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_6", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_6", "inbound_nodes": [[["upsample_1", 0, 0, {}], ["spatial_transformer_3", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_16", "trainable": true, "dtype": "float32", "output_dim": 640}, "name": "res_block_16", "inbound_nodes": [[["concatenate_6", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_10", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_10", "inbound_nodes": [[["res_block_16", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_7", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_7", "inbound_nodes": [[["spatial_transformer_10", 0, 0, {}], ["spatial_transformer_2", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_17", "trainable": true, "dtype": "float32", "output_dim": 640}, "name": "res_block_17", "inbound_nodes": [[["concatenate_7", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_11", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_11", "inbound_nodes": [[["res_block_17", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_8", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_8", "inbound_nodes": [[["spatial_transformer_11", 0, 0, {}], ["padded_conv2d_5", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_18", "trainable": true, "dtype": "float32", "output_dim": 640}, "name": "res_block_18", "inbound_nodes": [[["concatenate_8", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_12", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 10, "head_size": 64}, "name": "spatial_transformer_12", "inbound_nodes": [[["res_block_18", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Upsample", "config": {"name": "upsample_2", "trainable": true, "dtype": "float32", "channels": 640}, "name": "upsample_2", "inbound_nodes": [[["spatial_transformer_12", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_9", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_9", "inbound_nodes": [[["upsample_2", 0, 0, {}], ["spatial_transformer_1", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_19", "trainable": true, "dtype": "float32", "output_dim": 320}, "name": "res_block_19", "inbound_nodes": [[["concatenate_9", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_13", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_13", "inbound_nodes": [[["res_block_19", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_10", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_10", "inbound_nodes": [[["spatial_transformer_13", 0, 0, {}], ["spatial_transformer", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_20", "trainable": true, "dtype": "float32", "output_dim": 320}, "name": "res_block_20", "inbound_nodes": [[["concatenate_10", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_14", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_14", "inbound_nodes": [[["res_block_20", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "Concatenate", "config": {"name": "concatenate_11", "trainable": true, "dtype": "float32", "axis": -1}, "name": "concatenate_11", "inbound_nodes": [[["spatial_transformer_14", 0, 0, {}], ["padded_conv2d", 0, 0, {}]]]}, {"class_name": "ResBlock", "config": {"name": "res_block_21", "trainable": true, "dtype": "float32", "output_dim": 320}, "name": "res_block_21", "inbound_nodes": [[["concatenate_11", 0, 0, {}], ["dense_1", 0, 0, {}]]]}, {"class_name": "SpatialTransformer", "config": {"name": "spatial_transformer_15", "trainable": true, "dtype": "float32", "fully_connected": true, "num_heads": 5, "head_size": 64}, "name": "spatial_transformer_15", "inbound_nodes": [[["res_block_21", 0, 0, {}], ["input_1", 0, 0, {}]]]}, {"class_name": "GroupNormalization", "config": {"name": "group_normalization_60", "trainable": true, "dtype": "float32", "groups": 32, "axis": -1, "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}}, "gamma_initializer": {"class_name": "Ones", "config": {}}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "name": "group_normalization_60", "inbound_nodes": [[["spatial_transformer_15", 0, 0, {}]]]}, {"class_name": "Activation", "config": {"name": "activation_67", "trainable": true, "dtype": "float32", "activation": "swish"}, "name": "activation_67", "inbound_nodes": [[["group_normalization_60", 0, 0, {}]]]}, {"class_name": "PaddedConv2D", "config": {"name": "padded_conv2d_51", "trainable": true, "dtype": "float32", "kernel_size": 3, "padding": 1, "filters": 4}, "name": "padded_conv2d_51", "inbound_nodes": [[["activation_67", 0, 0, {}]]]}], "input_layers": [["input_3", 0, 0], ["input_2", 0, 0], ["input_1", 0, 0]], "output_layers": [["padded_conv2d_51", 0, 0]]}

fingerprint.pb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:96c680a428de6a10d5bc29a9beb9d9a2fff81ae86a6d74a785fdc74253de9a0d
-size 56
+oid sha256:0337af7483e6767081ce8e8cbd1f4d0fadd21d5ae3d9af020b4a8629b808a32f
+size 53

keras_metadata.pb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7df7f57bf82d403568c1b3fa62b409c77b446f7a9ee45304ef68c17e93b1a822
-size 796755
+oid sha256:cc51b48b4a6a340b56ca5288d62ed99088500090d9f6434104a31f597aff72d0
+size 679053

saved_model.pb

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f1ba1a7fca6ddd63f13f26214df87674057914dcd740b0e856172a332bd3bb5c
-size 21096177
+oid sha256:6f89ecbcc59f917148eac22d9883b35d49108764dc68d1b12ad66f357e1ed7f2
+size 19765801

variables/variables.data-00000-of-00001

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b2b92abb3eac36b8af79f2a237999a713ba34adc20f8c4e572086095f8448885
-size 3464158248
+oid sha256:c563a729624eec6cef239b870725a9333f90134822b3e7d43f74518ea279ca86
+size 3464156890