From 22521c79c63ab35dcd01566d5c7366ee650a8a6b Mon Sep 17 00:00:00 2001 From: Stefan Radev Date: Fri, 14 Feb 2025 09:24:39 -0500 Subject: [PATCH] Flexible Default Settings Across Inference Networks (#296) * Add default configs * Correct configs and nice defaults * use small network and fixed step integration for flow matching tests --------- Co-authored-by: larskue --- .../consistency_models/consistency_model.py | 20 +- .../couplings/single_coupling.py | 17 +- .../networks/flow_matching/flow_matching.py | 62 ++- .../networks/free_form_flow/free_form_flow.py | 36 +- bayesflow/networks/inference_network.py | 2 + examples/Two_Moons_Starter.ipynb | 496 +++++++++++------- tests/test_networks/conftest.py | 5 +- 7 files changed, 408 insertions(+), 230 deletions(-) diff --git a/bayesflow/networks/consistency_models/consistency_model.py b/bayesflow/networks/consistency_models/consistency_model.py index 03a27c6e6..b87f82936 100644 --- a/bayesflow/networks/consistency_models/consistency_model.py +++ b/bayesflow/networks/consistency_models/consistency_model.py @@ -29,6 +29,15 @@ class ConsistencyModel(InferenceNetwork): Discussion: https://openreview.net/forum?id=WNzy9bRDvG """ + MLP_DEFAULT_CONFIG = { + "widths": (256, 256, 256, 256, 256), + "activation": "mish", + "kernel_initializer": "he_normal", + "residual": True, + "dropout": 0.05, + "spectral_normalization": False, + } + def __init__( self, total_steps: int | float, @@ -65,12 +74,18 @@ def __init__( **kwargs : dict, optional, default: {} Additional keyword arguments """ - # Normal is the only supported base distribution for CMs super().__init__(base_distribution="normal", **keras_kwargs(kwargs)) self.total_steps = float(total_steps) - self.student = find_network(subnet, **kwargs.get("subnet_kwargs", {})) + if subnet == "mlp": + subnet_kwargs = ConsistencyModel.MLP_DEFAULT_CONFIG.copy() + subnet_kwargs.update(kwargs.get("subnet_kwargs", {})) + else: + subnet_kwargs = kwargs.get("subnet_kwargs", {}) + + self.student = find_network(subnet, **subnet_kwargs) + self.student_projector = keras.layers.Dense(units=None, bias_initializer="zeros", kernel_initializer="zeros") self.sigma2 = ops.convert_to_tensor(sigma2) @@ -82,6 +97,7 @@ def __init__( self.s0 = float(s0) self.s1 = float(s1) + # create variable that works with JIT compilation self.current_step = self.add_weight(name="current_step", initializer="zeros", trainable=False, dtype="int") self.current_step.assign(0) diff --git a/bayesflow/networks/coupling_flow/couplings/single_coupling.py b/bayesflow/networks/coupling_flow/couplings/single_coupling.py index eb7f65659..40661a5f8 100644 --- a/bayesflow/networks/coupling_flow/couplings/single_coupling.py +++ b/bayesflow/networks/coupling_flow/couplings/single_coupling.py @@ -15,10 +15,25 @@ class SingleCoupling(InvertibleLayer): Subnet output tensors are linearly mapped to the correct dimension. """ + MLP_DEFAULT_CONFIG = { + "widths": (128, 128), + "activation": "hard_silu", + "kernel_initializer": "glorot_uniform", + "residual": False, + "dropout": 0.05, + "spectral_normalization": False, + } + def __init__(self, subnet: str | type = "mlp", transform: str = "affine", **kwargs): super().__init__(**keras_kwargs(kwargs)) - self.network = find_network(subnet, **kwargs.get("subnet_kwargs", {})) + if subnet == "mlp": + subnet_kwargs = SingleCoupling.MLP_DEFAULT_CONFIG.copy() + subnet_kwargs.update(kwargs.get("subnet_kwargs", {})) + else: + subnet_kwargs = kwargs.get("subnet_kwargs", {}) + + self.network = find_network(subnet, **subnet_kwargs) self.transform = find_transform(transform, **kwargs.get("transform_kwargs", {})) output_projector_kwargs = kwargs.get("output_projector_kwargs", {}) diff --git a/bayesflow/networks/flow_matching/flow_matching.py b/bayesflow/networks/flow_matching/flow_matching.py index 07731edbe..79308822b 100644 --- a/bayesflow/networks/flow_matching/flow_matching.py +++ b/bayesflow/networks/flow_matching/flow_matching.py @@ -27,6 +27,31 @@ class FlowMatching(InferenceNetwork): [3] Optimal Transport Flow Matching: arXiv:2302.00482 """ + MLP_DEFAULT_CONFIG = { + "widths": (256, 256, 256, 256, 256), + "activation": "mish", + "kernel_initializer": "he_normal", + "residual": True, + "dropout": 0.05, + "spectral_normalization": False, + } + + OPTIMAL_TRANSPORT_DEFAULT_CONFIG = { + "method": "sinkhorn", + "cost": "euclidean", + "regularization": 0.1, + "max_steps": 100, + "tolerance": 1e-4, + } + + INTEGRATE_DEFAULT_CONFIG = { + "method": "rk45", + "steps": "adaptive", + "tolerance": 1e-3, + "min_steps": 10, + "max_steps": 100, + } + def __init__( self, subnet: str | type = "mlp", @@ -41,41 +66,28 @@ def __init__( self.use_optimal_transport = use_optimal_transport - if integrate_kwargs is None: - integrate_kwargs = { - "method": "rk45", - "steps": "adaptive", - "tolerance": 1e-3, - "min_steps": 10, - "max_steps": 100, - } - - self.integrate_kwargs = integrate_kwargs - - if optimal_transport_kwargs is None: - optimal_transport_kwargs = { - "method": "sinkhorn", - "cost": "euclidean", - "regularization": 0.1, - "max_steps": 100, - "tolerance": 1e-4, - } + self.integrate_kwargs = integrate_kwargs or FlowMatching.INTEGRATE_DEFAULT_CONFIG.copy() + self.optimal_transport_kwargs = optimal_transport_kwargs or FlowMatching.OPTIMAL_TRANSPORT_DEFAULT_CONFIG.copy() self.loss_fn = keras.losses.get(loss_fn) - self.optimal_transport_kwargs = optimal_transport_kwargs - self.seed_generator = keras.random.SeedGenerator() - self.subnet = find_network(subnet, **kwargs.get("subnet_kwargs", {})) + if subnet == "mlp": + subnet_kwargs = FlowMatching.MLP_DEFAULT_CONFIG.copy() + subnet_kwargs.update(kwargs.get("subnet_kwargs", {})) + else: + subnet_kwargs = kwargs.get("subnet_kwargs", {}) + + self.subnet = find_network(subnet, **subnet_kwargs) self.output_projector = keras.layers.Dense(units=None, bias_initializer="zeros") # serialization: store all parameters necessary to call __init__ self.config = { "base_distribution": base_distribution, - "use_optimal_transport": use_optimal_transport, - "optimal_transport_kwargs": optimal_transport_kwargs, - "integrate_kwargs": integrate_kwargs, + "use_optimal_transport": self.use_optimal_transport, + "optimal_transport_kwargs": self.optimal_transport_kwargs, + "integrate_kwargs": self.integrate_kwargs, **kwargs, } self.config = serialize_value_or_type(self.config, "subnet", subnet) diff --git a/bayesflow/networks/free_form_flow/free_form_flow.py b/bayesflow/networks/free_form_flow/free_form_flow.py index fd5ca180a..aa21dd63a 100644 --- a/bayesflow/networks/free_form_flow/free_form_flow.py +++ b/bayesflow/networks/free_form_flow/free_form_flow.py @@ -31,6 +31,24 @@ class FreeFormFlow(InferenceNetwork): In International Conference on Learning Representations. """ + ENCODER_MLP_DEFAULT_CONFIG = { + "widths": (256, 256, 256, 256), + "activation": "mish", + "kernel_initializer": "he_normal", + "residual": True, + "dropout": 0.05, + "spectral_normalization": False, + } + + DECODER_MLP_DEFAULT_CONFIG = { + "widths": (256, 256, 256, 256), + "activation": "mish", + "kernel_initializer": "he_normal", + "residual": True, + "dropout": 0.05, + "spectral_normalization": False, + } + def __init__( self, beta: float = 50.0, @@ -62,9 +80,23 @@ def __init__( Additional keyword arguments """ super().__init__(base_distribution=base_distribution, **keras_kwargs(kwargs)) - self.encoder_subnet = find_network(encoder_subnet, **kwargs.get("encoder_subnet_kwargs", {})) + + if encoder_subnet == "mlp": + encoder_subnet_kwargs = FreeFormFlow.ENCODER_MLP_DEFAULT_CONFIG.copy() + encoder_subnet_kwargs.update(kwargs.get("encoder_subnet_kwargs", {})) + else: + encoder_subnet_kwargs = kwargs.get("encoder_subnet_kwargs", {}) + + self.encoder_subnet = find_network(encoder_subnet, **encoder_subnet_kwargs) self.encoder_projector = keras.layers.Dense(units=None, bias_initializer="zeros", kernel_initializer="zeros") - self.decoder_subnet = find_network(decoder_subnet, **kwargs.get("decoder_subnet_kwargs", {})) + + if decoder_subnet == "mlp": + decoder_subnet_kwargs = FreeFormFlow.DECODER_MLP_DEFAULT_CONFIG.copy() + decoder_subnet_kwargs.update(kwargs.get("decoder_subnet_kwargs", {})) + else: + decoder_subnet_kwargs = kwargs.get("decoder_subnet_kwargs", {}) + + self.decoder_subnet = find_network(decoder_subnet, **decoder_subnet_kwargs) self.decoder_projector = keras.layers.Dense(units=None, bias_initializer="zeros", kernel_initializer="zeros") self.hutchinson_sampling = hutchinson_sampling diff --git a/bayesflow/networks/inference_network.py b/bayesflow/networks/inference_network.py index 868be5582..a637b7f0a 100644 --- a/bayesflow/networks/inference_network.py +++ b/bayesflow/networks/inference_network.py @@ -6,6 +6,8 @@ class InferenceNetwork(keras.Layer): + MLP_DEFAULT_CONFIG = {} + def __init__(self, base_distribution: str = "normal", **kwargs): super().__init__(**kwargs) self.base_distribution = find_distribution(base_distribution) diff --git a/examples/Two_Moons_Starter.ipynb b/examples/Two_Moons_Starter.ipynb index 5dc165ce4..2cdef5b84 100644 --- a/examples/Two_Moons_Starter.ipynb +++ b/examples/Two_Moons_Starter.ipynb @@ -25,7 +25,15 @@ "name": "stderr", "output_type": "stream", "text": [ - "INFO:bayesflow:Using backend 'tensorflow'\n" + "2025-02-11 16:22:30.912532: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n", + "2025-02-11 16:22:30.912881: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.\n", + "2025-02-11 16:22:30.914936: I external/local_tsl/tsl/cuda/cudart_stub.cc:32] Could not find cuda drivers on your machine, GPU will not be used.\n", + "2025-02-11 16:22:30.939014: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n", + "To enable the following instructions: AVX2 AVX_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n", + "2025-02-11 16:22:31.425109: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n", + "2025-02-11 16:22:31.935372: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:998] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n", + "2025-02-11 16:22:31.935638: W tensorflow/core/common_runtime/gpu/gpu_device.cc:2251] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.\n", + "Skipping registering GPU devices...\n" ] } ], @@ -150,7 +158,7 @@ "outputs": [], "source": [ "# generate 3 random draws from the joint distribution p(r, alpha, theta, x)\n", - "sample_data = simulator.sample((3,))" + "sample_data = simulator.sample(3)" ] }, { @@ -273,7 +281,7 @@ "num_training_batches = 512\n", "num_validation_batches = 128\n", "batch_size = 64\n", - "epochs = 30" + "epochs = 50" ] }, { @@ -322,7 +330,7 @@ "source": [ "flow_matching = bf.networks.FlowMatching(\n", " subnet=\"mlp\", \n", - " subnet_kwargs={\"widths\": (256,)*6 , \"dropout\": 0.0, \"residual\": True}\n", + " subnet_kwargs={\"dropout\": 0.0, \"widths\": (256,)*6} # override default dropout = 0.05 and widths = (256,)*5\n", ")" ] }, @@ -395,66 +403,106 @@ "name": "stdout", "output_type": "stream", "text": [ - "Epoch 1/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 4ms/step - loss: 0.4375 - loss/inference_loss: 0.4375 - val_loss: 0.3653 - val_loss/inference_loss: 0.3653\n", - "Epoch 2/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3750 - loss/inference_loss: 0.3750 - val_loss: 0.3006 - val_loss/inference_loss: 0.3006\n", - "Epoch 3/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3718 - loss/inference_loss: 0.3718 - val_loss: 0.4908 - val_loss/inference_loss: 0.4908\n", - "Epoch 4/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3625 - loss/inference_loss: 0.3625 - val_loss: 0.2568 - val_loss/inference_loss: 0.2568\n", - "Epoch 5/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3611 - loss/inference_loss: 0.3611 - val_loss: 0.3194 - val_loss/inference_loss: 0.3194\n", - "Epoch 6/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3545 - loss/inference_loss: 0.3545 - val_loss: 0.2798 - val_loss/inference_loss: 0.2798\n", - "Epoch 7/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3464 - loss/inference_loss: 0.3464 - val_loss: 0.3649 - val_loss/inference_loss: 0.3649\n", - "Epoch 8/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3376 - loss/inference_loss: 0.3376 - val_loss: 0.3246 - val_loss/inference_loss: 0.3246\n", - "Epoch 9/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3421 - loss/inference_loss: 0.3421 - val_loss: 0.3664 - val_loss/inference_loss: 0.3664\n", - "Epoch 10/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3361 - loss/inference_loss: 0.3361 - val_loss: 0.2294 - val_loss/inference_loss: 0.2294\n", - "Epoch 11/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3343 - loss/inference_loss: 0.3343 - val_loss: 0.3697 - val_loss/inference_loss: 0.3697\n", - "Epoch 12/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3378 - loss/inference_loss: 0.3378 - val_loss: 0.2370 - val_loss/inference_loss: 0.2370\n", - "Epoch 13/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3275 - loss/inference_loss: 0.3275 - val_loss: 0.2895 - val_loss/inference_loss: 0.2895\n", - "Epoch 14/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3358 - loss/inference_loss: 0.3358 - val_loss: 0.3811 - val_loss/inference_loss: 0.3811\n", - "Epoch 15/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3295 - loss/inference_loss: 0.3295 - val_loss: 0.3383 - val_loss/inference_loss: 0.3383\n", - "Epoch 16/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3293 - loss/inference_loss: 0.3293 - val_loss: 0.3162 - val_loss/inference_loss: 0.3162\n", - "Epoch 17/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3233 - loss/inference_loss: 0.3233 - val_loss: 0.5696 - val_loss/inference_loss: 0.5696\n", - "Epoch 18/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3248 - loss/inference_loss: 0.3248 - val_loss: 0.2916 - val_loss/inference_loss: 0.2916\n", - "Epoch 19/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3195 - loss/inference_loss: 0.3195 - val_loss: 0.3094 - val_loss/inference_loss: 0.3094\n", - "Epoch 20/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3219 - loss/inference_loss: 0.3219 - val_loss: 0.2837 - val_loss/inference_loss: 0.2837\n", - "Epoch 21/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3170 - loss/inference_loss: 0.3170 - val_loss: 0.1897 - val_loss/inference_loss: 0.1897\n", - "Epoch 22/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3178 - loss/inference_loss: 0.3178 - val_loss: 0.3624 - val_loss/inference_loss: 0.3624\n", - "Epoch 23/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3090 - loss/inference_loss: 0.3090 - val_loss: 0.5049 - val_loss/inference_loss: 0.5049\n", - "Epoch 24/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3108 - loss/inference_loss: 0.3108 - val_loss: 0.3213 - val_loss/inference_loss: 0.3213\n", - "Epoch 25/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3121 - loss/inference_loss: 0.3121 - val_loss: 0.3449 - val_loss/inference_loss: 0.3449\n", - "Epoch 26/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3158 - loss/inference_loss: 0.3158 - val_loss: 0.3167 - val_loss/inference_loss: 0.3167\n", - "Epoch 27/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3047 - loss/inference_loss: 0.3047 - val_loss: 0.2979 - val_loss/inference_loss: 0.2979\n", - "Epoch 28/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3170 - loss/inference_loss: 0.3170 - val_loss: 0.3634 - val_loss/inference_loss: 0.3634\n", - "Epoch 29/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3117 - loss/inference_loss: 0.3117 - val_loss: 0.4235 - val_loss/inference_loss: 0.4235\n", - "Epoch 30/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3115 - loss/inference_loss: 0.3115 - val_loss: 0.4121 - val_loss/inference_loss: 0.4121\n" + "Epoch 1/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - loss: 1.5975 - loss/inference_loss: 1.5975 - val_loss: 0.3615 - val_loss/inference_loss: 0.3615\n", + "Epoch 2/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3963 - loss/inference_loss: 0.3963 - val_loss: 0.3259 - val_loss/inference_loss: 0.3259\n", + "Epoch 3/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3793 - loss/inference_loss: 0.3793 - val_loss: 0.3774 - val_loss/inference_loss: 0.3774\n", + "Epoch 4/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3734 - loss/inference_loss: 0.3734 - val_loss: 0.2816 - val_loss/inference_loss: 0.2816\n", + "Epoch 5/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3625 - loss/inference_loss: 0.3625 - val_loss: 0.2620 - val_loss/inference_loss: 0.2620\n", + "Epoch 6/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3600 - loss/inference_loss: 0.3600 - val_loss: 0.2465 - val_loss/inference_loss: 0.2465\n", + "Epoch 7/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3646 - loss/inference_loss: 0.3646 - val_loss: 0.3418 - val_loss/inference_loss: 0.3418\n", + "Epoch 8/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3593 - loss/inference_loss: 0.3593 - val_loss: 0.3830 - val_loss/inference_loss: 0.3830\n", + "Epoch 9/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3493 - loss/inference_loss: 0.3493 - val_loss: 0.4498 - val_loss/inference_loss: 0.4498\n", + "Epoch 10/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3492 - loss/inference_loss: 0.3492 - val_loss: 0.3947 - val_loss/inference_loss: 0.3947\n", + "Epoch 11/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3459 - loss/inference_loss: 0.3459 - val_loss: 0.2970 - val_loss/inference_loss: 0.2970\n", + "Epoch 12/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3457 - loss/inference_loss: 0.3457 - val_loss: 0.2243 - val_loss/inference_loss: 0.2243\n", + "Epoch 13/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3451 - loss/inference_loss: 0.3451 - val_loss: 0.4153 - val_loss/inference_loss: 0.4153\n", + "Epoch 14/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3433 - loss/inference_loss: 0.3433 - val_loss: 0.3919 - val_loss/inference_loss: 0.3919\n", + "Epoch 15/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3373 - loss/inference_loss: 0.3373 - val_loss: 0.3400 - val_loss/inference_loss: 0.3400\n", + "Epoch 16/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3440 - loss/inference_loss: 0.3440 - val_loss: 0.1885 - val_loss/inference_loss: 0.1885\n", + "Epoch 17/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3429 - loss/inference_loss: 0.3429 - val_loss: 0.3297 - val_loss/inference_loss: 0.3297\n", + "Epoch 18/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3389 - loss/inference_loss: 0.3389 - val_loss: 0.3932 - val_loss/inference_loss: 0.3932\n", + "Epoch 19/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3330 - loss/inference_loss: 0.3330 - val_loss: 0.2640 - val_loss/inference_loss: 0.2640\n", + "Epoch 20/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3272 - loss/inference_loss: 0.3272 - val_loss: 0.3262 - val_loss/inference_loss: 0.3262\n", + "Epoch 21/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3301 - loss/inference_loss: 0.3301 - val_loss: 0.4301 - val_loss/inference_loss: 0.4301\n", + "Epoch 22/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3329 - loss/inference_loss: 0.3329 - val_loss: 0.4407 - val_loss/inference_loss: 0.4407\n", + "Epoch 23/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3384 - loss/inference_loss: 0.3384 - val_loss: 0.2786 - val_loss/inference_loss: 0.2786\n", + "Epoch 24/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3283 - loss/inference_loss: 0.3283 - val_loss: 0.3840 - val_loss/inference_loss: 0.3840\n", + "Epoch 25/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3235 - loss/inference_loss: 0.3235 - val_loss: 0.3168 - val_loss/inference_loss: 0.3168\n", + "Epoch 26/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3227 - loss/inference_loss: 0.3227 - val_loss: 0.2289 - val_loss/inference_loss: 0.2289\n", + "Epoch 27/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3367 - loss/inference_loss: 0.3367 - val_loss: 0.2283 - val_loss/inference_loss: 0.2283\n", + "Epoch 28/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3331 - loss/inference_loss: 0.3331 - val_loss: 0.3331 - val_loss/inference_loss: 0.3331\n", + "Epoch 29/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3281 - loss/inference_loss: 0.3281 - val_loss: 0.1447 - val_loss/inference_loss: 0.1447\n", + "Epoch 30/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3228 - loss/inference_loss: 0.3228 - val_loss: 0.2868 - val_loss/inference_loss: 0.2868\n", + "Epoch 31/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3226 - loss/inference_loss: 0.3226 - val_loss: 0.2819 - val_loss/inference_loss: 0.2819\n", + "Epoch 32/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3167 - loss/inference_loss: 0.3167 - val_loss: 0.3676 - val_loss/inference_loss: 0.3676\n", + "Epoch 33/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3264 - loss/inference_loss: 0.3264 - val_loss: 0.2303 - val_loss/inference_loss: 0.2303\n", + "Epoch 34/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3141 - loss/inference_loss: 0.3141 - val_loss: 0.2125 - val_loss/inference_loss: 0.2125\n", + "Epoch 35/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3096 - loss/inference_loss: 0.3096 - val_loss: 0.2754 - val_loss/inference_loss: 0.2754\n", + "Epoch 36/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3187 - loss/inference_loss: 0.3187 - val_loss: 0.3006 - val_loss/inference_loss: 0.3006\n", + "Epoch 37/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3157 - loss/inference_loss: 0.3157 - val_loss: 0.3113 - val_loss/inference_loss: 0.3113\n", + "Epoch 38/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3212 - loss/inference_loss: 0.3212 - val_loss: 0.4190 - val_loss/inference_loss: 0.4190\n", + "Epoch 39/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3162 - loss/inference_loss: 0.3162 - val_loss: 0.3351 - val_loss/inference_loss: 0.3351\n", + "Epoch 40/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3159 - loss/inference_loss: 0.3159 - val_loss: 0.3813 - val_loss/inference_loss: 0.3813\n", + "Epoch 41/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3134 - loss/inference_loss: 0.3134 - val_loss: 0.4158 - val_loss/inference_loss: 0.4158\n", + "Epoch 42/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3142 - loss/inference_loss: 0.3142 - val_loss: 0.5772 - val_loss/inference_loss: 0.5772\n", + "Epoch 43/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3136 - loss/inference_loss: 0.3136 - val_loss: 0.3419 - val_loss/inference_loss: 0.3419\n", + "Epoch 44/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 3ms/step - loss: 0.3104 - loss/inference_loss: 0.3104 - val_loss: 0.5761 - val_loss/inference_loss: 0.5761\n", + "Epoch 45/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3166 - loss/inference_loss: 0.3166 - val_loss: 0.2678 - val_loss/inference_loss: 0.2678\n", + "Epoch 46/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3197 - loss/inference_loss: 0.3197 - val_loss: 0.3576 - val_loss/inference_loss: 0.3576\n", + "Epoch 47/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3165 - loss/inference_loss: 0.3165 - val_loss: 0.3637 - val_loss/inference_loss: 0.3637\n", + "Epoch 48/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3145 - loss/inference_loss: 0.3145 - val_loss: 0.3348 - val_loss/inference_loss: 0.3348\n", + "Epoch 49/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3104 - loss/inference_loss: 0.3104 - val_loss: 0.3068 - val_loss/inference_loss: 0.3068\n", + "Epoch 50/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3090 - loss/inference_loss: 0.3090 - val_loss: 0.4407 - val_loss/inference_loss: 0.4407\n" ] } ], @@ -548,7 +596,7 @@ "# Compute the empirical variance of the draws from the prior θ ~ p(θ)\n", "consistency_model = bf.networks.ConsistencyModel(\n", " subnet=\"mlp\",\n", - " subnet_kwargs={\"widths\": (256,)*6, \"dropout\": 0.0, \"residual\": True},\n", + " subnet_kwargs={\"dropout\": 0.0, \"widths\": (256,)*6},\n", " total_steps=num_training_batches*epochs,\n", " max_time=10, # this probably needs to be tuned for a novel application\n", " sigma2=1.0, # the data adapter standardizes our parameters, so set to 1.0\n", @@ -588,66 +636,106 @@ "name": "stdout", "output_type": "stream", "text": [ - "Epoch 1/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 4ms/step - loss: 0.3610 - loss/inference_loss: 0.3610 - val_loss: 0.3100 - val_loss/inference_loss: 0.3100\n", - "Epoch 2/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.3014 - loss/inference_loss: 0.3014 - val_loss: 0.3638 - val_loss/inference_loss: 0.3638\n", - "Epoch 3/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2846 - loss/inference_loss: 0.2846 - val_loss: 0.3233 - val_loss/inference_loss: 0.3233\n", - "Epoch 4/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2767 - loss/inference_loss: 0.2767 - val_loss: 0.3025 - val_loss/inference_loss: 0.3025\n", - "Epoch 5/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2683 - loss/inference_loss: 0.2683 - val_loss: 0.3097 - val_loss/inference_loss: 0.3097\n", - "Epoch 6/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2674 - loss/inference_loss: 0.2674 - val_loss: 0.2850 - val_loss/inference_loss: 0.2850\n", - "Epoch 7/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2625 - loss/inference_loss: 0.2625 - val_loss: 0.1652 - val_loss/inference_loss: 0.1652\n", - "Epoch 8/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2542 - loss/inference_loss: 0.2542 - val_loss: 0.2975 - val_loss/inference_loss: 0.2975\n", - "Epoch 9/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2552 - loss/inference_loss: 0.2552 - val_loss: 0.2748 - val_loss/inference_loss: 0.2748\n", - "Epoch 10/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2697 - loss/inference_loss: 0.2697 - val_loss: 0.3729 - val_loss/inference_loss: 0.3729\n", - "Epoch 11/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2616 - loss/inference_loss: 0.2616 - val_loss: 0.1787 - val_loss/inference_loss: 0.1787\n", - "Epoch 12/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2558 - loss/inference_loss: 0.2558 - val_loss: 0.2838 - val_loss/inference_loss: 0.2838\n", - "Epoch 13/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2504 - loss/inference_loss: 0.2504 - val_loss: 0.2136 - val_loss/inference_loss: 0.2136\n", - "Epoch 14/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2560 - loss/inference_loss: 0.2560 - val_loss: 0.2751 - val_loss/inference_loss: 0.2751\n", - "Epoch 15/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2438 - loss/inference_loss: 0.2438 - val_loss: 0.2211 - val_loss/inference_loss: 0.2211\n", - "Epoch 16/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2448 - loss/inference_loss: 0.2448 - val_loss: 0.1773 - val_loss/inference_loss: 0.1773\n", - "Epoch 17/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2391 - loss/inference_loss: 0.2391 - val_loss: 0.3960 - val_loss/inference_loss: 0.3960\n", - "Epoch 18/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2334 - loss/inference_loss: 0.2334 - val_loss: 0.1638 - val_loss/inference_loss: 0.1638\n", - "Epoch 19/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2388 - loss/inference_loss: 0.2388 - val_loss: 0.4104 - val_loss/inference_loss: 0.4104\n", - "Epoch 20/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2274 - loss/inference_loss: 0.2274 - val_loss: 0.1807 - val_loss/inference_loss: 0.1807\n", - "Epoch 21/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2228 - loss/inference_loss: 0.2228 - val_loss: 0.1728 - val_loss/inference_loss: 0.1728\n", - "Epoch 22/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2288 - loss/inference_loss: 0.2288 - val_loss: 0.1764 - val_loss/inference_loss: 0.1764\n", - "Epoch 23/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2237 - loss/inference_loss: 0.2237 - val_loss: 0.2428 - val_loss/inference_loss: 0.2428\n", - "Epoch 24/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2197 - loss/inference_loss: 0.2197 - val_loss: 0.1803 - val_loss/inference_loss: 0.1803\n", - "Epoch 25/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: 0.2233 - loss/inference_loss: 0.2233 - val_loss: 0.3102 - val_loss/inference_loss: 0.3102\n", - "Epoch 26/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: 0.2257 - loss/inference_loss: 0.2257 - val_loss: 0.1427 - val_loss/inference_loss: 0.1427\n", - "Epoch 27/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: 0.2132 - loss/inference_loss: 0.2132 - val_loss: 0.4167 - val_loss/inference_loss: 0.4167\n", - "Epoch 28/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2077 - loss/inference_loss: 0.2077 - val_loss: 0.1379 - val_loss/inference_loss: 0.1379\n", - "Epoch 29/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2143 - loss/inference_loss: 0.2143 - val_loss: 0.1349 - val_loss/inference_loss: 0.1349\n", - "Epoch 30/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2084 - loss/inference_loss: 0.2084 - val_loss: 0.1330 - val_loss/inference_loss: 0.1330\n" + "Epoch 1/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 4ms/step - loss: 0.3549 - loss/inference_loss: 0.3549 - val_loss: 0.2793 - val_loss/inference_loss: 0.2793\n", + "Epoch 2/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.3020 - loss/inference_loss: 0.3020 - val_loss: 0.3505 - val_loss/inference_loss: 0.3505\n", + "Epoch 3/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2831 - loss/inference_loss: 0.2831 - val_loss: 0.2429 - val_loss/inference_loss: 0.2429\n", + "Epoch 4/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2813 - loss/inference_loss: 0.2813 - val_loss: 0.3600 - val_loss/inference_loss: 0.3600\n", + "Epoch 5/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2732 - loss/inference_loss: 0.2732 - val_loss: 0.2537 - val_loss/inference_loss: 0.2537\n", + "Epoch 6/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2722 - loss/inference_loss: 0.2722 - val_loss: 0.2904 - val_loss/inference_loss: 0.2904\n", + "Epoch 7/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2618 - loss/inference_loss: 0.2618 - val_loss: 0.1984 - val_loss/inference_loss: 0.1984\n", + "Epoch 8/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2663 - loss/inference_loss: 0.2663 - val_loss: 0.1680 - val_loss/inference_loss: 0.1680\n", + "Epoch 9/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2599 - loss/inference_loss: 0.2599 - val_loss: 0.2595 - val_loss/inference_loss: 0.2595\n", + "Epoch 10/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2567 - loss/inference_loss: 0.2567 - val_loss: 0.2612 - val_loss/inference_loss: 0.2612\n", + "Epoch 11/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2530 - loss/inference_loss: 0.2530 - val_loss: 0.2694 - val_loss/inference_loss: 0.2694\n", + "Epoch 12/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2596 - loss/inference_loss: 0.2596 - val_loss: 0.3073 - val_loss/inference_loss: 0.3073\n", + "Epoch 13/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2566 - loss/inference_loss: 0.2566 - val_loss: 0.1798 - val_loss/inference_loss: 0.1798\n", + "Epoch 14/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2589 - loss/inference_loss: 0.2589 - val_loss: 0.2743 - val_loss/inference_loss: 0.2743\n", + "Epoch 15/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2493 - loss/inference_loss: 0.2493 - val_loss: 0.2189 - val_loss/inference_loss: 0.2189\n", + "Epoch 16/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2649 - loss/inference_loss: 0.2649 - val_loss: 0.2154 - val_loss/inference_loss: 0.2154\n", + "Epoch 17/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2609 - loss/inference_loss: 0.2609 - val_loss: 0.2758 - val_loss/inference_loss: 0.2758\n", + "Epoch 18/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2526 - loss/inference_loss: 0.2526 - val_loss: 0.1542 - val_loss/inference_loss: 0.1542\n", + "Epoch 19/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2510 - loss/inference_loss: 0.2510 - val_loss: 0.1860 - val_loss/inference_loss: 0.1860\n", + "Epoch 20/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2559 - loss/inference_loss: 0.2559 - val_loss: 0.2213 - val_loss/inference_loss: 0.2213\n", + "Epoch 21/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2524 - loss/inference_loss: 0.2524 - val_loss: 0.2497 - val_loss/inference_loss: 0.2497\n", + "Epoch 22/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2477 - loss/inference_loss: 0.2477 - val_loss: 0.2030 - val_loss/inference_loss: 0.2030\n", + "Epoch 23/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2447 - loss/inference_loss: 0.2447 - val_loss: 0.2862 - val_loss/inference_loss: 0.2862\n", + "Epoch 24/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2451 - loss/inference_loss: 0.2451 - val_loss: 0.3859 - val_loss/inference_loss: 0.3859\n", + "Epoch 25/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2429 - loss/inference_loss: 0.2429 - val_loss: 0.2310 - val_loss/inference_loss: 0.2310\n", + "Epoch 26/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2437 - loss/inference_loss: 0.2437 - val_loss: 0.2236 - val_loss/inference_loss: 0.2236\n", + "Epoch 27/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2432 - loss/inference_loss: 0.2432 - val_loss: 0.3466 - val_loss/inference_loss: 0.3466\n", + "Epoch 28/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2353 - loss/inference_loss: 0.2353 - val_loss: 0.2234 - val_loss/inference_loss: 0.2234\n", + "Epoch 29/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2351 - loss/inference_loss: 0.2351 - val_loss: 0.1637 - val_loss/inference_loss: 0.1637\n", + "Epoch 30/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2364 - loss/inference_loss: 0.2364 - val_loss: 0.2324 - val_loss/inference_loss: 0.2324\n", + "Epoch 31/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2310 - loss/inference_loss: 0.2310 - val_loss: 0.1853 - val_loss/inference_loss: 0.1853\n", + "Epoch 32/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2287 - loss/inference_loss: 0.2287 - val_loss: 0.1234 - val_loss/inference_loss: 0.1234\n", + "Epoch 33/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2159 - loss/inference_loss: 0.2159 - val_loss: 0.1985 - val_loss/inference_loss: 0.1985\n", + "Epoch 34/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2228 - loss/inference_loss: 0.2228 - val_loss: 0.4063 - val_loss/inference_loss: 0.4063\n", + "Epoch 35/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2155 - loss/inference_loss: 0.2155 - val_loss: 0.2233 - val_loss/inference_loss: 0.2233\n", + "Epoch 36/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2257 - loss/inference_loss: 0.2257 - val_loss: 0.1208 - val_loss/inference_loss: 0.1208\n", + "Epoch 37/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2231 - loss/inference_loss: 0.2231 - val_loss: 0.0776 - val_loss/inference_loss: 0.0776\n", + "Epoch 38/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2193 - loss/inference_loss: 0.2193 - val_loss: 0.2310 - val_loss/inference_loss: 0.2310\n", + "Epoch 39/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2204 - loss/inference_loss: 0.2204 - val_loss: 0.1733 - val_loss/inference_loss: 0.1733\n", + "Epoch 40/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2219 - loss/inference_loss: 0.2219 - val_loss: 0.1291 - val_loss/inference_loss: 0.1291\n", + "Epoch 41/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2127 - loss/inference_loss: 0.2127 - val_loss: 0.1073 - val_loss/inference_loss: 0.1073\n", + "Epoch 42/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2257 - loss/inference_loss: 0.2257 - val_loss: 0.2174 - val_loss/inference_loss: 0.2174\n", + "Epoch 43/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2178 - loss/inference_loss: 0.2178 - val_loss: 0.2001 - val_loss/inference_loss: 0.2001\n", + "Epoch 44/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2150 - loss/inference_loss: 0.2150 - val_loss: 0.2282 - val_loss/inference_loss: 0.2282\n", + "Epoch 45/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2130 - loss/inference_loss: 0.2130 - val_loss: 0.1956 - val_loss/inference_loss: 0.1956\n", + "Epoch 46/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2046 - loss/inference_loss: 0.2046 - val_loss: 0.1937 - val_loss/inference_loss: 0.1937\n", + "Epoch 47/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2082 - loss/inference_loss: 0.2082 - val_loss: 0.2303 - val_loss/inference_loss: 0.2303\n", + "Epoch 48/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2067 - loss/inference_loss: 0.2067 - val_loss: 0.1300 - val_loss/inference_loss: 0.1300\n", + "Epoch 49/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: 0.2201 - loss/inference_loss: 0.2201 - val_loss: 0.1432 - val_loss/inference_loss: 0.1432\n", + "Epoch 50/50\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 3ms/step - loss: 0.2097 - loss/inference_loss: 0.2097 - val_loss: 0.1791 - val_loss/inference_loss: 0.1791\n" ] } ], @@ -672,19 +760,29 @@ }, { "cell_type": "code", - "execution_count": 24, + "execution_count": 14, "id": "27b83a8f", "metadata": {}, "outputs": [], "source": [ "affine_flow = bf.networks.CouplingFlow(subnet=\"mlp\")\n", "\n", - "spline_flow = bf.networks.CouplingFlow(subnet=\"mlp\", transform=\"spline\")" + "spline_flow = bf.networks.CouplingFlow(subnet=\"mlp\", transform=\"spline\", depth=4)" ] }, { "cell_type": "code", - "execution_count": 25, + "execution_count": 15, + "id": "73b7a64d", + "metadata": {}, + "outputs": [], + "source": [ + "epochs = 30 # coupling flows need less epochs than free-form methods" + ] + }, + { + "cell_type": "code", + "execution_count": 16, "id": "e634dc50", "metadata": {}, "outputs": [], @@ -713,7 +811,7 @@ }, { "cell_type": "code", - "execution_count": 26, + "execution_count": 17, "id": "f52e8e49", "metadata": {}, "outputs": [ @@ -730,72 +828,72 @@ "output_type": "stream", "text": [ "Epoch 1/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m24s\u001b[0m 17ms/step - loss: -1.7248 - loss/inference_loss: -1.7248 - val_loss: -2.1969 - val_loss/inference_loss: -2.1969\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 6ms/step - loss: -1.2724 - loss/inference_loss: -1.2724 - val_loss: -2.2869 - val_loss/inference_loss: -2.2869\n", "Epoch 2/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m8s\u001b[0m 16ms/step - loss: -2.2365 - loss/inference_loss: -2.2365 - val_loss: -2.3102 - val_loss/inference_loss: -2.3102\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -2.4998 - loss/inference_loss: -2.4998 - val_loss: -2.1505 - val_loss/inference_loss: -2.1505\n", "Epoch 3/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m9s\u001b[0m 17ms/step - loss: -2.3881 - loss/inference_loss: -2.3881 - val_loss: -2.3807 - val_loss/inference_loss: -2.3807\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -2.6645 - loss/inference_loss: -2.6645 - val_loss: -2.6893 - val_loss/inference_loss: -2.6893\n", "Epoch 4/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -2.3680 - loss/inference_loss: -2.3680 - val_loss: -2.1068 - val_loss/inference_loss: -2.1068\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: -2.7494 - loss/inference_loss: -2.7494 - val_loss: -2.4783 - val_loss/inference_loss: -2.4783\n", "Epoch 5/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.5437 - loss/inference_loss: -2.5437 - val_loss: -2.5328 - val_loss/inference_loss: -2.5328\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -2.7852 - loss/inference_loss: -2.7852 - val_loss: -2.7409 - val_loss/inference_loss: -2.7409\n", "Epoch 6/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.6639 - loss/inference_loss: -2.6639 - val_loss: -2.6180 - val_loss/inference_loss: -2.6180\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: -2.8575 - loss/inference_loss: -2.8575 - val_loss: -2.6789 - val_loss/inference_loss: -2.6789\n", "Epoch 7/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -2.7511 - loss/inference_loss: -2.7511 - val_loss: -2.7643 - val_loss/inference_loss: -2.7643\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: -2.9127 - loss/inference_loss: -2.9127 - val_loss: -3.0435 - val_loss/inference_loss: -3.0435\n", "Epoch 8/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.8458 - loss/inference_loss: -2.8458 - val_loss: -2.6873 - val_loss/inference_loss: -2.6873\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -2.9346 - loss/inference_loss: -2.9346 - val_loss: -2.9628 - val_loss/inference_loss: -2.9628\n", "Epoch 9/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.9340 - loss/inference_loss: -2.9340 - val_loss: -3.0405 - val_loss/inference_loss: -3.0405\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -2.9371 - loss/inference_loss: -2.9371 - val_loss: -2.3899 - val_loss/inference_loss: -2.3899\n", "Epoch 10/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.9189 - loss/inference_loss: -2.9189 - val_loss: -2.8785 - val_loss/inference_loss: -2.8785\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 4ms/step - loss: -2.9999 - loss/inference_loss: -2.9999 - val_loss: -2.9758 - val_loss/inference_loss: -2.9758\n", "Epoch 11/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -2.9837 - loss/inference_loss: -2.9837 - val_loss: -2.7903 - val_loss/inference_loss: -2.7903\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.0083 - loss/inference_loss: -3.0083 - val_loss: -2.5975 - val_loss/inference_loss: -2.5975\n", "Epoch 12/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.0446 - loss/inference_loss: -3.0446 - val_loss: -2.9181 - val_loss/inference_loss: -2.9181\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.0434 - loss/inference_loss: -3.0434 - val_loss: -3.0387 - val_loss/inference_loss: -3.0387\n", "Epoch 13/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -3.0572 - loss/inference_loss: -3.0572 - val_loss: -3.1326 - val_loss/inference_loss: -3.1326\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.0920 - loss/inference_loss: -3.0920 - val_loss: -2.3253 - val_loss/inference_loss: -2.3253\n", "Epoch 14/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -3.1098 - loss/inference_loss: -3.1098 - val_loss: -2.8643 - val_loss/inference_loss: -2.8643\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.1255 - loss/inference_loss: -3.1255 - val_loss: -3.1556 - val_loss/inference_loss: -3.1556\n", "Epoch 15/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.1765 - loss/inference_loss: -3.1765 - val_loss: -2.9744 - val_loss/inference_loss: -2.9744\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.1654 - loss/inference_loss: -3.1654 - val_loss: -2.8726 - val_loss/inference_loss: -2.8726\n", "Epoch 16/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.2259 - loss/inference_loss: -3.2259 - val_loss: -3.2496 - val_loss/inference_loss: -3.2496\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.1779 - loss/inference_loss: -3.1779 - val_loss: -3.1917 - val_loss/inference_loss: -3.1917\n", "Epoch 17/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.2587 - loss/inference_loss: -3.2587 - val_loss: -3.2098 - val_loss/inference_loss: -3.2098\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.2124 - loss/inference_loss: -3.2124 - val_loss: -2.1857 - val_loss/inference_loss: -2.1857\n", "Epoch 18/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.3191 - loss/inference_loss: -3.3191 - val_loss: -3.4182 - val_loss/inference_loss: -3.4182\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.2409 - loss/inference_loss: -3.2409 - val_loss: -2.9640 - val_loss/inference_loss: -2.9640\n", "Epoch 19/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.3424 - loss/inference_loss: -3.3424 - val_loss: -3.2258 - val_loss/inference_loss: -3.2258\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.2689 - loss/inference_loss: -3.2689 - val_loss: -2.7462 - val_loss/inference_loss: -2.7462\n", "Epoch 20/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.3740 - loss/inference_loss: -3.3740 - val_loss: -3.3169 - val_loss/inference_loss: -3.3169\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.2759 - loss/inference_loss: -3.2759 - val_loss: -2.9262 - val_loss/inference_loss: -2.9262\n", "Epoch 21/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -3.4080 - loss/inference_loss: -3.4080 - val_loss: -3.3350 - val_loss/inference_loss: -3.3350\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.3185 - loss/inference_loss: -3.3185 - val_loss: -2.1308 - val_loss/inference_loss: -2.1308\n", "Epoch 22/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -3.4475 - loss/inference_loss: -3.4475 - val_loss: -3.3964 - val_loss/inference_loss: -3.3964\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.3447 - loss/inference_loss: -3.3447 - val_loss: -2.6838 - val_loss/inference_loss: -2.6838\n", "Epoch 23/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.4809 - loss/inference_loss: -3.4809 - val_loss: -3.3064 - val_loss/inference_loss: -3.3064\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.3607 - loss/inference_loss: -3.3607 - val_loss: -3.4517 - val_loss/inference_loss: -3.4517\n", "Epoch 24/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.5171 - loss/inference_loss: -3.5171 - val_loss: -3.2936 - val_loss/inference_loss: -3.2936\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.3540 - loss/inference_loss: -3.3540 - val_loss: -3.0357 - val_loss/inference_loss: -3.0357\n", "Epoch 25/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.5438 - loss/inference_loss: -3.5438 - val_loss: -3.3020 - val_loss/inference_loss: -3.3020\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.3831 - loss/inference_loss: -3.3831 - val_loss: -3.3768 - val_loss/inference_loss: -3.3768\n", "Epoch 26/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.5721 - loss/inference_loss: -3.5721 - val_loss: -3.6407 - val_loss/inference_loss: -3.6407\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.4183 - loss/inference_loss: -3.4183 - val_loss: -3.1671 - val_loss/inference_loss: -3.1671\n", "Epoch 27/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.5948 - loss/inference_loss: -3.5948 - val_loss: -3.6379 - val_loss/inference_loss: -3.6379\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.4194 - loss/inference_loss: -3.4194 - val_loss: 9.2951 - val_loss/inference_loss: 9.2951\n", "Epoch 28/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 13ms/step - loss: -3.6117 - loss/inference_loss: -3.6117 - val_loss: -3.3497 - val_loss/inference_loss: -3.3497\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.3851 - loss/inference_loss: -3.3851 - val_loss: -2.9896 - val_loss/inference_loss: -2.9896\n", "Epoch 29/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.6138 - loss/inference_loss: -3.6138 - val_loss: -3.5269 - val_loss/inference_loss: -3.5269\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 5ms/step - loss: -3.4251 - loss/inference_loss: -3.4251 - val_loss: -3.3175 - val_loss/inference_loss: -3.3175\n", "Epoch 30/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m7s\u001b[0m 14ms/step - loss: -3.5773 - loss/inference_loss: -3.5773 - val_loss: -3.4148 - val_loss/inference_loss: -3.4148\n" + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 5ms/step - loss: -3.4083 - loss/inference_loss: -3.4083 - val_loss: -3.2599 - val_loss/inference_loss: -3.2599\n" ] } ], "source": [ "history = affine_flow_workflow.fit_offline(\n", " training_data, \n", - " epochs=epochs, \n", + " epochs=epochs,\n", " batch_size=batch_size,\n", " validation_data=validation_data\n", ")" @@ -803,7 +901,7 @@ }, { "cell_type": "code", - "execution_count": 27, + "execution_count": 18, "id": "afa9839f", "metadata": {}, "outputs": [ @@ -820,72 +918,72 @@ "output_type": "stream", "text": [ "Epoch 1/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m50s\u001b[0m 31ms/step - loss: -0.0305 - loss/inference_loss: -0.0305 - val_loss: -0.5046 - val_loss/inference_loss: -0.5046\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 11ms/step - loss: -1.0211 - loss/inference_loss: -1.0211 - val_loss: -1.5653 - val_loss/inference_loss: -1.5653\n", "Epoch 2/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -1.2209 - loss/inference_loss: -1.2209 - val_loss: -1.4572 - val_loss/inference_loss: -1.4572\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.2318 - loss/inference_loss: -2.2318 - val_loss: -2.4503 - val_loss/inference_loss: -2.4503\n", "Epoch 3/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -1.5152 - loss/inference_loss: -1.5152 - val_loss: -1.5478 - val_loss/inference_loss: -1.5478\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.5659 - loss/inference_loss: -2.5659 - val_loss: -2.6406 - val_loss/inference_loss: -2.6406\n", "Epoch 4/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -1.7783 - loss/inference_loss: -1.7783 - val_loss: -1.6943 - val_loss/inference_loss: -1.6943\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.7533 - loss/inference_loss: -2.7533 - val_loss: -2.6080 - val_loss/inference_loss: -2.6080\n", "Epoch 5/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -1.9830 - loss/inference_loss: -1.9830 - val_loss: -2.3109 - val_loss/inference_loss: -2.3109\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.8607 - loss/inference_loss: -2.8607 - val_loss: -2.8657 - val_loss/inference_loss: -2.8657\n", "Epoch 6/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -1.9812 - loss/inference_loss: -1.9812 - val_loss: -2.6307 - val_loss/inference_loss: -2.6307\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.9422 - loss/inference_loss: -2.9422 - val_loss: -2.3686 - val_loss/inference_loss: -2.3686\n", "Epoch 7/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.2906 - loss/inference_loss: -2.2906 - val_loss: -2.0684 - val_loss/inference_loss: -2.0684\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -2.9989 - loss/inference_loss: -2.9989 - val_loss: -2.9271 - val_loss/inference_loss: -2.9271\n", "Epoch 8/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 25ms/step - loss: -2.3631 - loss/inference_loss: -2.3631 - val_loss: -2.2277 - val_loss/inference_loss: -2.2277\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.0638 - loss/inference_loss: -3.0638 - val_loss: -3.0360 - val_loss/inference_loss: -3.0360\n", "Epoch 9/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.3808 - loss/inference_loss: -2.3808 - val_loss: -2.4927 - val_loss/inference_loss: -2.4927\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.0553 - loss/inference_loss: -3.0553 - val_loss: -3.2254 - val_loss/inference_loss: -3.2254\n", "Epoch 10/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.5112 - loss/inference_loss: -2.5112 - val_loss: -2.6694 - val_loss/inference_loss: -2.6694\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.1095 - loss/inference_loss: -3.1095 - val_loss: -3.0538 - val_loss/inference_loss: -3.0538\n", "Epoch 11/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -2.5705 - loss/inference_loss: -2.5705 - val_loss: -2.7786 - val_loss/inference_loss: -2.7786\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.2038 - loss/inference_loss: -3.2038 - val_loss: -3.1451 - val_loss/inference_loss: -3.1451\n", "Epoch 12/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -2.6979 - loss/inference_loss: -2.6979 - val_loss: -2.6825 - val_loss/inference_loss: -2.6825\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.2396 - loss/inference_loss: -3.2396 - val_loss: -3.2923 - val_loss/inference_loss: -3.2923\n", "Epoch 13/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.8183 - loss/inference_loss: -2.8183 - val_loss: -2.7877 - val_loss/inference_loss: -2.7877\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.2695 - loss/inference_loss: -3.2695 - val_loss: -2.7734 - val_loss/inference_loss: -2.7734\n", "Epoch 14/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.7840 - loss/inference_loss: -2.7840 - val_loss: -2.8402 - val_loss/inference_loss: -2.8402\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.2778 - loss/inference_loss: -3.2778 - val_loss: -3.3034 - val_loss/inference_loss: -3.3034\n", "Epoch 15/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -2.3101 - loss/inference_loss: -2.3101 - val_loss: -2.8259 - val_loss/inference_loss: -2.8259\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.3281 - loss/inference_loss: -3.3281 - val_loss: -2.5565 - val_loss/inference_loss: -2.5565\n", "Epoch 16/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -2.9274 - loss/inference_loss: -2.9274 - val_loss: -2.9762 - val_loss/inference_loss: -2.9762\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.3416 - loss/inference_loss: -3.3416 - val_loss: -3.1074 - val_loss/inference_loss: -3.1074\n", "Epoch 17/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.0012 - loss/inference_loss: -3.0012 - val_loss: -2.6167 - val_loss/inference_loss: -2.6167\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.3612 - loss/inference_loss: -3.3612 - val_loss: -3.4145 - val_loss/inference_loss: -3.4145\n", "Epoch 18/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -3.0712 - loss/inference_loss: -3.0712 - val_loss: -3.1585 - val_loss/inference_loss: -3.1585\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.3806 - loss/inference_loss: -3.3806 - val_loss: -3.3677 - val_loss/inference_loss: -3.3677\n", "Epoch 19/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -3.1087 - loss/inference_loss: -3.1087 - val_loss: -2.7591 - val_loss/inference_loss: -2.7591\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.4344 - loss/inference_loss: -3.4344 - val_loss: -2.7881 - val_loss/inference_loss: -2.7881\n", "Epoch 20/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.1919 - loss/inference_loss: -3.1919 - val_loss: -3.3845 - val_loss/inference_loss: -3.3845\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.4364 - loss/inference_loss: -3.4364 - val_loss: -3.2963 - val_loss/inference_loss: -3.2963\n", "Epoch 21/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -3.2309 - loss/inference_loss: -3.2309 - val_loss: -3.4535 - val_loss/inference_loss: -3.4535\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.4700 - loss/inference_loss: -3.4700 - val_loss: -2.8206 - val_loss/inference_loss: -2.8206\n", "Epoch 22/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.3305 - loss/inference_loss: -3.3305 - val_loss: -3.1565 - val_loss/inference_loss: -3.1565\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.4920 - loss/inference_loss: -3.4920 - val_loss: -3.0479 - val_loss/inference_loss: -3.0479\n", "Epoch 23/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.3760 - loss/inference_loss: -3.3760 - val_loss: -3.3958 - val_loss/inference_loss: -3.3958\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5153 - loss/inference_loss: -3.5153 - val_loss: -3.0690 - val_loss/inference_loss: -3.0690\n", "Epoch 24/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -3.4144 - loss/inference_loss: -3.4144 - val_loss: -3.2900 - val_loss/inference_loss: -3.2900\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5309 - loss/inference_loss: -3.5309 - val_loss: -2.9115 - val_loss/inference_loss: -2.9115\n", "Epoch 25/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.4843 - loss/inference_loss: -3.4843 - val_loss: -3.3188 - val_loss/inference_loss: -3.3188\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5516 - loss/inference_loss: -3.5516 - val_loss: -3.0176 - val_loss/inference_loss: -3.0176\n", "Epoch 26/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.5172 - loss/inference_loss: -3.5172 - val_loss: -3.5762 - val_loss/inference_loss: -3.5762\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5659 - loss/inference_loss: -3.5659 - val_loss: -3.0676 - val_loss/inference_loss: -3.0676\n", "Epoch 27/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.5435 - loss/inference_loss: -3.5435 - val_loss: -3.4907 - val_loss/inference_loss: -3.4907\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5792 - loss/inference_loss: -3.5792 - val_loss: -3.3448 - val_loss/inference_loss: -3.3448\n", "Epoch 28/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.5667 - loss/inference_loss: -3.5667 - val_loss: -3.4892 - val_loss/inference_loss: -3.4892\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5791 - loss/inference_loss: -3.5791 - val_loss: -3.1989 - val_loss/inference_loss: -3.1989\n", "Epoch 29/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 23ms/step - loss: -3.5856 - loss/inference_loss: -3.5856 - val_loss: -3.2707 - val_loss/inference_loss: -3.2707\n", + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.5712 - loss/inference_loss: -3.5712 - val_loss: -3.0113 - val_loss/inference_loss: -3.0113\n", "Epoch 30/30\n", - "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 24ms/step - loss: -3.5624 - loss/inference_loss: -3.5624 - val_loss: -3.4194 - val_loss/inference_loss: -3.4194\n" + "\u001b[1m512/512\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 8ms/step - loss: -3.4733 - loss/inference_loss: -3.4733 - val_loss: -3.4334 - val_loss/inference_loss: -3.4334\n" ] } ], "source": [ "history = spline_flow_workflow.fit_offline(\n", - " training_data, \n", - " epochs=epochs, \n", + " training_data,\n", + " epochs=epochs,\n", " batch_size=batch_size,\n", " validation_data=validation_data\n", ")" @@ -913,13 +1011,13 @@ }, { "cell_type": "code", - "execution_count": 28, + "execution_count": 19, "id": "073bcd0b", "metadata": {}, "outputs": [ { "data": { - "image/png": "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", + "image/png": "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", "text/plain": [ "
" ] @@ -983,8 +1081,8 @@ }, { "cell_type": "code", - "execution_count": 37, - "id": "df35a911", + "execution_count": null, + "id": "886bce5c", "metadata": {}, "outputs": [], "source": [] @@ -992,7 +1090,7 @@ ], "metadata": { "kernelspec": { - "display_name": "bf", + "display_name": "bayeskeras", "language": "python", "name": "python3" }, @@ -1006,7 +1104,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.11.5" + "version": "3.11.4" }, "toc": { "base_numbering": 1, diff --git a/tests/test_networks/conftest.py b/tests/test_networks/conftest.py index d8eac634c..a42289f34 100644 --- a/tests/test_networks/conftest.py +++ b/tests/test_networks/conftest.py @@ -28,7 +28,10 @@ def subnet(request): def flow_matching(): from bayesflow.networks import FlowMatching - return FlowMatching() + return FlowMatching( + subnet_kwargs={"widths": None, "width": 64, "depth": 2}, + integrate_kwargs={"method": "rk45", "steps": 100}, + ) @pytest.fixture()