Note: This environment variable is required for fully deterministic CuBLAS ops on CUDA >= 10.2 when
reproducible=Trueis set below. Without it, PyTorch raises aRuntimeErrorinstead of training deterministically. It must be set beforetorchis imported. See the README FAQ for details.
In [ ]:
Copied!
%env CUBLAS_WORKSPACE_CONFIG=:16:8
%env CUBLAS_WORKSPACE_CONFIG=:16:8
Evaluate Overview¶
This tutorial provides an overview of the in-built functionalities to evaluate the learned embeddings for downstream machine learning tasks leveraging metadata information.
Evaluation is designed as additional pipeline step after .run() (which includes preprocessing, training, predict on test data and visualization).
When you work with a pre-trained model, make sure that you perform the .predict() step such that embeddings are calculated for newly provided test-samples.
Stackix example¶
We will use a Stackix as an example to show options in embedding evaluation.
In [1]:
Copied!
# Train the Stackix similar to pipeline tutorial
import autoencodix as acx
from autoencodix.configs.default_config import DataCase
from autoencodix.configs.stackix_config import StackixConfig
from autoencodix.utils.example_data import EXAMPLE_MULTI_BULK
# Stackix has its own config class
# instead of passing a pandas DataFrame, we use a pre-defined DataPackage object directly.
# this time with single cell data
print("Input data:")
print(EXAMPLE_MULTI_BULK)
print("-" * 50)
my_config = StackixConfig(
epochs=100,
checkpoint_interval=5,
default_vae_loss="kl",
data_case=DataCase.MULTI_BULK,
)
print("\n")
print("Starting Pipeline")
print("-" * 50)
print("-" * 50)
stackix = acx.Stackix(data=EXAMPLE_MULTI_BULK, config=my_config)
result = stackix.run()
# Train the Stackix similar to pipeline tutorial
import autoencodix as acx
from autoencodix.configs.default_config import DataCase
from autoencodix.configs.stackix_config import StackixConfig
from autoencodix.utils.example_data import EXAMPLE_MULTI_BULK
# Stackix has its own config class
# instead of passing a pandas DataFrame, we use a pre-defined DataPackage object directly.
# this time with single cell data
print("Input data:")
print(EXAMPLE_MULTI_BULK)
print("-" * 50)
my_config = StackixConfig(
epochs=100,
checkpoint_interval=5,
default_vae_loss="kl",
data_case=DataCase.MULTI_BULK,
)
print("\n")
print("Starting Pipeline")
print("-" * 50)
print("-" * 50)
stackix = acx.Stackix(data=EXAMPLE_MULTI_BULK, config=my_config)
result = stackix.run()
Input data: multi_bulk: transcriptomics: 500 samples × 100 features proteomics: 500 samples × 80 features annotation: transcriptomics: 500 samples × 3 features proteomics: 500 samples × 3 features -------------------------------------------------- Starting Pipeline -------------------------------------------------- -------------------------------------------------- in handle_direct_user_data with data: <class 'autoencodix.data.datapackage.DataPackage'> anno key: transcriptomics anno key: proteomics Training each modality model... Training modality: transcriptomics Training modality: transcriptomics Epoch 1 - Train Loss: 119.1194 Sub-losses: recon_loss: 119.1194, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 1 - Valid Loss: 105.4678 Sub-losses: recon_loss: 105.4678, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 2 - Train Loss: 114.0099 Sub-losses: recon_loss: 114.0099, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 2 - Valid Loss: 106.6898 Sub-losses: recon_loss: 106.6898, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Train Loss: 111.4732 Sub-losses: recon_loss: 111.4732, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Valid Loss: 106.3609 Sub-losses: recon_loss: 106.3609, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Train Loss: 109.4193 Sub-losses: recon_loss: 109.4193, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Valid Loss: 107.2363 Sub-losses: recon_loss: 107.2363, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Train Loss: 107.9533 Sub-losses: recon_loss: 107.9532, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Valid Loss: 107.1992 Sub-losses: recon_loss: 107.1992, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Train Loss: 107.0810 Sub-losses: recon_loss: 107.0809, var_loss: 0.0001, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Valid Loss: 106.7958 Sub-losses: recon_loss: 106.7958, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 7 - Train Loss: 106.3023 Sub-losses: recon_loss: 106.3022, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 7 - Valid Loss: 103.8173 Sub-losses: recon_loss: 103.8172, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Train Loss: 105.0296 Sub-losses: recon_loss: 105.0295, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Valid Loss: 103.7357 Sub-losses: recon_loss: 103.7356, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Train Loss: 104.2397 Sub-losses: recon_loss: 104.2396, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Valid Loss: 103.9822 Sub-losses: recon_loss: 103.9821, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 10 - Train Loss: 104.4244 Sub-losses: recon_loss: 104.4243, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 10 - Valid Loss: 103.1529 Sub-losses: recon_loss: 103.1528, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Train Loss: 103.3126 Sub-losses: recon_loss: 103.3124, var_loss: 0.0002, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Valid Loss: 103.1606 Sub-losses: recon_loss: 103.1605, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 12 - Train Loss: 103.1324 Sub-losses: recon_loss: 103.1322, var_loss: 0.0002, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 12 - Valid Loss: 102.4310 Sub-losses: recon_loss: 102.4309, var_loss: 0.0001, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 13 - Train Loss: 102.4150 Sub-losses: recon_loss: 102.4148, var_loss: 0.0002, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 13 - Valid Loss: 102.2809 Sub-losses: recon_loss: 102.2807, var_loss: 0.0002, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 14 - Train Loss: 101.9913 Sub-losses: recon_loss: 101.9910, var_loss: 0.0003, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 14 - Valid Loss: 102.4041 Sub-losses: recon_loss: 102.4038, var_loss: 0.0002, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 15 - Train Loss: 101.3868 Sub-losses: recon_loss: 101.3864, var_loss: 0.0004, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 15 - Valid Loss: 101.6762 Sub-losses: recon_loss: 101.6759, var_loss: 0.0003, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 16 - Train Loss: 100.5582 Sub-losses: recon_loss: 100.5576, var_loss: 0.0006, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 16 - Valid Loss: 100.9372 Sub-losses: recon_loss: 100.9368, var_loss: 0.0004, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 17 - Train Loss: 100.1677 Sub-losses: recon_loss: 100.1670, var_loss: 0.0007, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 17 - Valid Loss: 100.0215 Sub-losses: recon_loss: 100.0210, var_loss: 0.0005, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 18 - Train Loss: 98.7768 Sub-losses: recon_loss: 98.7759, var_loss: 0.0009, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 18 - Valid Loss: 99.5905 Sub-losses: recon_loss: 99.5898, var_loss: 0.0006, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 19 - Train Loss: 98.1422 Sub-losses: recon_loss: 98.1409, var_loss: 0.0013, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 19 - Valid Loss: 98.8752 Sub-losses: recon_loss: 98.8744, var_loss: 0.0009, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 20 - Train Loss: 97.9796 Sub-losses: recon_loss: 97.9780, var_loss: 0.0016, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 20 - Valid Loss: 98.5548 Sub-losses: recon_loss: 98.5535, var_loss: 0.0013, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 21 - Train Loss: 96.6108 Sub-losses: recon_loss: 96.6087, var_loss: 0.0020, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 21 - Valid Loss: 98.0430 Sub-losses: recon_loss: 98.0415, var_loss: 0.0015, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 22 - Train Loss: 96.7423 Sub-losses: recon_loss: 96.7396, var_loss: 0.0027, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 22 - Valid Loss: 98.1283 Sub-losses: recon_loss: 98.1264, var_loss: 0.0020, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 23 - Train Loss: 95.4074 Sub-losses: recon_loss: 95.4040, var_loss: 0.0035, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 23 - Valid Loss: 98.0078 Sub-losses: recon_loss: 98.0052, var_loss: 0.0025, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 24 - Train Loss: 95.1056 Sub-losses: recon_loss: 95.1012, var_loss: 0.0044, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 24 - Valid Loss: 94.2064 Sub-losses: recon_loss: 94.2032, var_loss: 0.0032, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 25 - Train Loss: 94.2514 Sub-losses: recon_loss: 94.2458, var_loss: 0.0055, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 25 - Valid Loss: 94.5971 Sub-losses: recon_loss: 94.5928, var_loss: 0.0043, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 26 - Train Loss: 94.6982 Sub-losses: recon_loss: 94.6908, var_loss: 0.0075, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 26 - Valid Loss: 94.3439 Sub-losses: recon_loss: 94.3381, var_loss: 0.0058, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 27 - Train Loss: 93.0567 Sub-losses: recon_loss: 93.0475, var_loss: 0.0092, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 27 - Valid Loss: 92.6010 Sub-losses: recon_loss: 92.5937, var_loss: 0.0073, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 28 - Train Loss: 92.7516 Sub-losses: recon_loss: 92.7403, var_loss: 0.0113, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 28 - Valid Loss: 93.3897 Sub-losses: recon_loss: 93.3802, var_loss: 0.0096, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 29 - Train Loss: 92.9076 Sub-losses: recon_loss: 92.8930, var_loss: 0.0146, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 29 - Valid Loss: 90.7657 Sub-losses: recon_loss: 90.7532, var_loss: 0.0125, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 30 - Train Loss: 92.3022 Sub-losses: recon_loss: 92.2835, var_loss: 0.0187, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 30 - Valid Loss: 90.0650 Sub-losses: recon_loss: 90.0488, var_loss: 0.0163, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 31 - Train Loss: 91.3403 Sub-losses: recon_loss: 91.3157, var_loss: 0.0246, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 31 - Valid Loss: 89.9511 Sub-losses: recon_loss: 89.9293, var_loss: 0.0218, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 32 - Train Loss: 90.9510 Sub-losses: recon_loss: 90.9215, var_loss: 0.0295, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 32 - Valid Loss: 89.2959 Sub-losses: recon_loss: 89.2682, var_loss: 0.0276, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 33 - Train Loss: 90.2779 Sub-losses: recon_loss: 90.2406, var_loss: 0.0373, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 33 - Valid Loss: 90.3272 Sub-losses: recon_loss: 90.2909, var_loss: 0.0362, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 34 - Train Loss: 90.4713 Sub-losses: recon_loss: 90.4227, var_loss: 0.0486, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 34 - Valid Loss: 86.4606 Sub-losses: recon_loss: 86.4111, var_loss: 0.0496, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 35 - Train Loss: 89.4123 Sub-losses: recon_loss: 89.3509, var_loss: 0.0614, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 35 - Valid Loss: 86.2724 Sub-losses: recon_loss: 86.2059, var_loss: 0.0666, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 36 - Train Loss: 89.2934 Sub-losses: recon_loss: 89.2140, var_loss: 0.0794, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 36 - Valid Loss: 84.9099 Sub-losses: recon_loss: 84.8304, var_loss: 0.0795, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 37 - Train Loss: 88.5790 Sub-losses: recon_loss: 88.4793, var_loss: 0.0998, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 37 - Valid Loss: 85.0665 Sub-losses: recon_loss: 84.9662, var_loss: 0.1004, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 38 - Train Loss: 88.7223 Sub-losses: recon_loss: 88.5993, var_loss: 0.1229, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 38 - Valid Loss: 84.1812 Sub-losses: recon_loss: 84.0564, var_loss: 0.1248, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 39 - Train Loss: 88.5020 Sub-losses: recon_loss: 88.3523, var_loss: 0.1497, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 39 - Valid Loss: 85.9974 Sub-losses: recon_loss: 85.8429, var_loss: 0.1545, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 40 - Train Loss: 87.7164 Sub-losses: recon_loss: 87.5270, var_loss: 0.1894, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 40 - Valid Loss: 84.3068 Sub-losses: recon_loss: 84.1200, var_loss: 0.1868, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 41 - Train Loss: 86.6250 Sub-losses: recon_loss: 86.3922, var_loss: 0.2328, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 41 - Valid Loss: 85.8306 Sub-losses: recon_loss: 85.6135, var_loss: 0.2171, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 42 - Train Loss: 87.0210 Sub-losses: recon_loss: 86.7450, var_loss: 0.2760, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 42 - Valid Loss: 84.5616 Sub-losses: recon_loss: 84.2870, var_loss: 0.2745, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 43 - Train Loss: 87.1063 Sub-losses: recon_loss: 86.7773, var_loss: 0.3290, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 43 - Valid Loss: 83.3701 Sub-losses: recon_loss: 83.0308, var_loss: 0.3393, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 44 - Train Loss: 87.2573 Sub-losses: recon_loss: 86.8450, var_loss: 0.4122, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 44 - Valid Loss: 83.3995 Sub-losses: recon_loss: 82.9749, var_loss: 0.4246, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 45 - Train Loss: 85.6736 Sub-losses: recon_loss: 85.1832, var_loss: 0.4904, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 45 - Valid Loss: 84.7959 Sub-losses: recon_loss: 84.2828, var_loss: 0.5131, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 46 - Train Loss: 86.2236 Sub-losses: recon_loss: 85.6295, var_loss: 0.5941, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 46 - Valid Loss: 83.2876 Sub-losses: recon_loss: 82.6647, var_loss: 0.6229, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 47 - Train Loss: 85.7799 Sub-losses: recon_loss: 85.0819, var_loss: 0.6980, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 47 - Valid Loss: 83.8451 Sub-losses: recon_loss: 83.1457, var_loss: 0.6994, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 48 - Train Loss: 85.7284 Sub-losses: recon_loss: 84.9203, var_loss: 0.8081, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 48 - Valid Loss: 82.9488 Sub-losses: recon_loss: 82.1214, var_loss: 0.8274, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 49 - Train Loss: 86.0633 Sub-losses: recon_loss: 85.1747, var_loss: 0.8886, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 49 - Valid Loss: 82.5187 Sub-losses: recon_loss: 81.6085, var_loss: 0.9102, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 50 - Train Loss: 86.2389 Sub-losses: recon_loss: 85.2157, var_loss: 1.0232, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 50 - Valid Loss: 82.2951 Sub-losses: recon_loss: 81.2732, var_loss: 1.0220, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 51 - Train Loss: 86.8495 Sub-losses: recon_loss: 85.6858, var_loss: 1.1637, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 51 - Valid Loss: 82.6324 Sub-losses: recon_loss: 81.4692, var_loss: 1.1631, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 52 - Train Loss: 86.6278 Sub-losses: recon_loss: 85.3513, var_loss: 1.2765, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 52 - Valid Loss: 82.7640 Sub-losses: recon_loss: 81.4701, var_loss: 1.2939, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 53 - Train Loss: 86.0967 Sub-losses: recon_loss: 84.6975, var_loss: 1.3992, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 53 - Valid Loss: 84.2376 Sub-losses: recon_loss: 82.9049, var_loss: 1.3327, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 54 - Train Loss: 86.0632 Sub-losses: recon_loss: 84.5720, var_loss: 1.4912, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 54 - Valid Loss: 83.6010 Sub-losses: recon_loss: 82.1732, var_loss: 1.4278, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 55 - Train Loss: 85.2406 Sub-losses: recon_loss: 83.7047, var_loss: 1.5359, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 55 - Valid Loss: 82.8333 Sub-losses: recon_loss: 81.3051, var_loss: 1.5282, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 56 - Train Loss: 85.5093 Sub-losses: recon_loss: 83.8809, var_loss: 1.6284, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 56 - Valid Loss: 82.8166 Sub-losses: recon_loss: 81.2958, var_loss: 1.5208, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 57 - Train Loss: 85.7360 Sub-losses: recon_loss: 84.0744, var_loss: 1.6616, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 57 - Valid Loss: 83.2285 Sub-losses: recon_loss: 81.6494, var_loss: 1.5791, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 58 - Train Loss: 86.1532 Sub-losses: recon_loss: 84.4577, var_loss: 1.6955, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 58 - Valid Loss: 83.9543 Sub-losses: recon_loss: 82.3178, var_loss: 1.6365, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 59 - Train Loss: 85.5716 Sub-losses: recon_loss: 83.8258, var_loss: 1.7459, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 59 - Valid Loss: 82.5119 Sub-losses: recon_loss: 80.7840, var_loss: 1.7280, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 60 - Train Loss: 85.5723 Sub-losses: recon_loss: 83.7650, var_loss: 1.8074, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 60 - Valid Loss: 82.4761 Sub-losses: recon_loss: 80.6573, var_loss: 1.8189, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 61 - Train Loss: 85.1828 Sub-losses: recon_loss: 83.3901, var_loss: 1.7927, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 61 - Valid Loss: 82.7882 Sub-losses: recon_loss: 80.9705, var_loss: 1.8178, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 62 - Train Loss: 85.1202 Sub-losses: recon_loss: 83.2570, var_loss: 1.8633, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 62 - Valid Loss: 81.9293 Sub-losses: recon_loss: 80.1283, var_loss: 1.8010, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 63 - Train Loss: 85.3766 Sub-losses: recon_loss: 83.5075, var_loss: 1.8691, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 63 - Valid Loss: 82.6217 Sub-losses: recon_loss: 80.8381, var_loss: 1.7836, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 64 - Train Loss: 85.7294 Sub-losses: recon_loss: 83.9316, var_loss: 1.7978, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 64 - Valid Loss: 82.6017 Sub-losses: recon_loss: 80.7418, var_loss: 1.8599, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 65 - Train Loss: 84.4224 Sub-losses: recon_loss: 82.5266, var_loss: 1.8958, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 65 - Valid Loss: 82.8188 Sub-losses: recon_loss: 80.9925, var_loss: 1.8263, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 66 - Train Loss: 84.8748 Sub-losses: recon_loss: 82.9977, var_loss: 1.8771, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 66 - Valid Loss: 81.9535 Sub-losses: recon_loss: 80.1761, var_loss: 1.7774, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 67 - Train Loss: 85.3699 Sub-losses: recon_loss: 83.4446, var_loss: 1.9254, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 67 - Valid Loss: 82.3956 Sub-losses: recon_loss: 80.6388, var_loss: 1.7568, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 68 - Train Loss: 84.5753 Sub-losses: recon_loss: 82.7211, var_loss: 1.8542, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 68 - Valid Loss: 81.1735 Sub-losses: recon_loss: 79.4058, var_loss: 1.7678, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 69 - Train Loss: 85.2817 Sub-losses: recon_loss: 83.4104, var_loss: 1.8713, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 69 - Valid Loss: 81.6565 Sub-losses: recon_loss: 79.8295, var_loss: 1.8270, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 70 - Train Loss: 84.3648 Sub-losses: recon_loss: 82.5027, var_loss: 1.8621, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 70 - Valid Loss: 81.3026 Sub-losses: recon_loss: 79.4844, var_loss: 1.8182, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 71 - Train Loss: 83.5772 Sub-losses: recon_loss: 81.6923, var_loss: 1.8849, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 71 - Valid Loss: 81.7928 Sub-losses: recon_loss: 79.9659, var_loss: 1.8268, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 72 - Train Loss: 84.3019 Sub-losses: recon_loss: 82.4779, var_loss: 1.8240, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 72 - Valid Loss: 81.0886 Sub-losses: recon_loss: 79.3001, var_loss: 1.7885, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 73 - Train Loss: 84.0004 Sub-losses: recon_loss: 82.1360, var_loss: 1.8645, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 73 - Valid Loss: 81.6669 Sub-losses: recon_loss: 79.8681, var_loss: 1.7987, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 74 - Train Loss: 83.8428 Sub-losses: recon_loss: 81.9931, var_loss: 1.8497, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 74 - Valid Loss: 81.5151 Sub-losses: recon_loss: 79.7349, var_loss: 1.7802, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 75 - Train Loss: 82.7972 Sub-losses: recon_loss: 80.9290, var_loss: 1.8683, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 75 - Valid Loss: 81.1549 Sub-losses: recon_loss: 79.3568, var_loss: 1.7981, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 76 - Train Loss: 83.1562 Sub-losses: recon_loss: 81.2664, var_loss: 1.8898, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 76 - Valid Loss: 80.7845 Sub-losses: recon_loss: 79.0287, var_loss: 1.7558, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 77 - Train Loss: 83.4703 Sub-losses: recon_loss: 81.5834, var_loss: 1.8870, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 77 - Valid Loss: 81.1933 Sub-losses: recon_loss: 79.4719, var_loss: 1.7214, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 78 - Train Loss: 83.0747 Sub-losses: recon_loss: 81.2221, var_loss: 1.8526, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 78 - Valid Loss: 81.0590 Sub-losses: recon_loss: 79.3883, var_loss: 1.6707, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 79 - Train Loss: 83.6625 Sub-losses: recon_loss: 81.8668, var_loss: 1.7957, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 79 - Valid Loss: 80.6059 Sub-losses: recon_loss: 78.8773, var_loss: 1.7285, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 80 - Train Loss: 83.1492 Sub-losses: recon_loss: 81.3099, var_loss: 1.8393, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 80 - Valid Loss: 80.7146 Sub-losses: recon_loss: 78.9619, var_loss: 1.7527, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 81 - Train Loss: 82.7625 Sub-losses: recon_loss: 80.9006, var_loss: 1.8619, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 81 - Valid Loss: 81.1412 Sub-losses: recon_loss: 79.4025, var_loss: 1.7388, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 82 - Train Loss: 84.0181 Sub-losses: recon_loss: 82.1011, var_loss: 1.9169, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 82 - Valid Loss: 80.7175 Sub-losses: recon_loss: 79.0044, var_loss: 1.7131, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 83 - Train Loss: 83.3030 Sub-losses: recon_loss: 81.4884, var_loss: 1.8145, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 83 - Valid Loss: 81.7071 Sub-losses: recon_loss: 80.0058, var_loss: 1.7013, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 84 - Train Loss: 83.1791 Sub-losses: recon_loss: 81.3482, var_loss: 1.8309, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 84 - Valid Loss: 81.3707 Sub-losses: recon_loss: 79.5963, var_loss: 1.7744, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 85 - Train Loss: 83.1488 Sub-losses: recon_loss: 81.3194, var_loss: 1.8294, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 85 - Valid Loss: 80.7035 Sub-losses: recon_loss: 78.9835, var_loss: 1.7199, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 86 - Train Loss: 83.2615 Sub-losses: recon_loss: 81.3742, var_loss: 1.8873, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 86 - Valid Loss: 82.0300 Sub-losses: recon_loss: 80.3164, var_loss: 1.7136, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 87 - Train Loss: 83.2846 Sub-losses: recon_loss: 81.4081, var_loss: 1.8766, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 87 - Valid Loss: 81.6411 Sub-losses: recon_loss: 79.9530, var_loss: 1.6881, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 88 - Train Loss: 83.2720 Sub-losses: recon_loss: 81.4224, var_loss: 1.8496, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 88 - Valid Loss: 82.2998 Sub-losses: recon_loss: 80.6060, var_loss: 1.6938, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 89 - Train Loss: 82.4550 Sub-losses: recon_loss: 80.5628, var_loss: 1.8922, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 89 - Valid Loss: 80.7147 Sub-losses: recon_loss: 78.9770, var_loss: 1.7377, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 90 - Train Loss: 82.7799 Sub-losses: recon_loss: 80.9001, var_loss: 1.8798, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 90 - Valid Loss: 81.2864 Sub-losses: recon_loss: 79.5300, var_loss: 1.7564, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 91 - Train Loss: 82.4761 Sub-losses: recon_loss: 80.5856, var_loss: 1.8905, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 91 - Valid Loss: 81.5463 Sub-losses: recon_loss: 79.7642, var_loss: 1.7821, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Train Loss: 82.6773 Sub-losses: recon_loss: 80.7910, var_loss: 1.8863, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Valid Loss: 81.6817 Sub-losses: recon_loss: 79.9466, var_loss: 1.7351, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 93 - Train Loss: 82.6559 Sub-losses: recon_loss: 80.7673, var_loss: 1.8886, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 93 - Valid Loss: 80.9823 Sub-losses: recon_loss: 79.2917, var_loss: 1.6907, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Train Loss: 83.5025 Sub-losses: recon_loss: 81.6214, var_loss: 1.8810, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Valid Loss: 80.9622 Sub-losses: recon_loss: 79.1978, var_loss: 1.7645, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Train Loss: 82.2063 Sub-losses: recon_loss: 80.3336, var_loss: 1.8727, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Valid Loss: 80.7530 Sub-losses: recon_loss: 78.9589, var_loss: 1.7941, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 96 - Train Loss: 81.7838 Sub-losses: recon_loss: 79.8523, var_loss: 1.9315, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 96 - Valid Loss: 81.8821 Sub-losses: recon_loss: 80.1365, var_loss: 1.7456, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Train Loss: 82.1349 Sub-losses: recon_loss: 80.2493, var_loss: 1.8856, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Valid Loss: 80.7317 Sub-losses: recon_loss: 79.0178, var_loss: 1.7138, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Train Loss: 81.7410 Sub-losses: recon_loss: 79.8139, var_loss: 1.9272, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Valid Loss: 80.2914 Sub-losses: recon_loss: 78.5774, var_loss: 1.7140, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Train Loss: 82.2451 Sub-losses: recon_loss: 80.3596, var_loss: 1.8855, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Valid Loss: 81.9952 Sub-losses: recon_loss: 80.2860, var_loss: 1.7092, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Train Loss: 82.2792 Sub-losses: recon_loss: 80.3596, var_loss: 1.9196, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Valid Loss: 80.5175 Sub-losses: recon_loss: 78.7193, var_loss: 1.7981, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Training modality: proteomics Training modality: proteomics Epoch 1 - Train Loss: 96.7905 Sub-losses: recon_loss: 96.7905, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 1 - Valid Loss: 84.7973 Sub-losses: recon_loss: 84.7973, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 2 - Train Loss: 93.5819 Sub-losses: recon_loss: 93.5819, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 2 - Valid Loss: 84.2610 Sub-losses: recon_loss: 84.2610, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Train Loss: 90.5553 Sub-losses: recon_loss: 90.5552, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Valid Loss: 85.9100 Sub-losses: recon_loss: 85.9100, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Train Loss: 88.9539 Sub-losses: recon_loss: 88.9539, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Valid Loss: 84.7894 Sub-losses: recon_loss: 84.7894, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Train Loss: 87.3968 Sub-losses: recon_loss: 87.3967, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Valid Loss: 84.2111 Sub-losses: recon_loss: 84.2110, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Train Loss: 86.9564 Sub-losses: recon_loss: 86.9563, var_loss: 0.0001, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Valid Loss: 85.2791 Sub-losses: recon_loss: 85.2791, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 7 - Train Loss: 85.8824 Sub-losses: recon_loss: 85.8824, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 7 - Valid Loss: 84.5515 Sub-losses: recon_loss: 84.5515, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Train Loss: 85.3596 Sub-losses: recon_loss: 85.3595, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Valid Loss: 83.6958 Sub-losses: recon_loss: 83.6958, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Train Loss: 84.2307 Sub-losses: recon_loss: 84.2306, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Valid Loss: 82.7694 Sub-losses: recon_loss: 82.7694, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 10 - Train Loss: 84.2398 Sub-losses: recon_loss: 84.2396, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 10 - Valid Loss: 84.2630 Sub-losses: recon_loss: 84.2629, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Train Loss: 83.2951 Sub-losses: recon_loss: 83.2950, var_loss: 0.0002, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Valid Loss: 80.9829 Sub-losses: recon_loss: 80.9828, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 12 - Train Loss: 82.9829 Sub-losses: recon_loss: 82.9827, var_loss: 0.0002, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 12 - Valid Loss: 82.0151 Sub-losses: recon_loss: 82.0150, var_loss: 0.0002, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 13 - Train Loss: 81.9261 Sub-losses: recon_loss: 81.9258, var_loss: 0.0003, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 13 - Valid Loss: 81.9222 Sub-losses: recon_loss: 81.9220, var_loss: 0.0002, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 14 - Train Loss: 81.6396 Sub-losses: recon_loss: 81.6393, var_loss: 0.0003, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 14 - Valid Loss: 81.4215 Sub-losses: recon_loss: 81.4213, var_loss: 0.0003, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 15 - Train Loss: 81.8077 Sub-losses: recon_loss: 81.8073, var_loss: 0.0004, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 15 - Valid Loss: 82.1703 Sub-losses: recon_loss: 82.1699, var_loss: 0.0003, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 16 - Train Loss: 81.5002 Sub-losses: recon_loss: 81.4996, var_loss: 0.0006, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 16 - Valid Loss: 81.9555 Sub-losses: recon_loss: 81.9552, var_loss: 0.0004, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 17 - Train Loss: 81.0463 Sub-losses: recon_loss: 81.0456, var_loss: 0.0007, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 17 - Valid Loss: 80.8832 Sub-losses: recon_loss: 80.8827, var_loss: 0.0005, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 18 - Train Loss: 80.3241 Sub-losses: recon_loss: 80.3231, var_loss: 0.0010, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 18 - Valid Loss: 80.0630 Sub-losses: recon_loss: 80.0624, var_loss: 0.0007, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 19 - Train Loss: 79.8691 Sub-losses: recon_loss: 79.8680, var_loss: 0.0012, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 19 - Valid Loss: 81.1680 Sub-losses: recon_loss: 81.1671, var_loss: 0.0008, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 20 - Train Loss: 79.8624 Sub-losses: recon_loss: 79.8609, var_loss: 0.0015, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 20 - Valid Loss: 79.8067 Sub-losses: recon_loss: 79.8056, var_loss: 0.0011, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 21 - Train Loss: 79.5141 Sub-losses: recon_loss: 79.5118, var_loss: 0.0023, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 21 - Valid Loss: 80.2688 Sub-losses: recon_loss: 80.2673, var_loss: 0.0015, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 22 - Train Loss: 78.7335 Sub-losses: recon_loss: 78.7306, var_loss: 0.0029, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 22 - Valid Loss: 79.3316 Sub-losses: recon_loss: 79.3295, var_loss: 0.0021, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 23 - Train Loss: 77.8730 Sub-losses: recon_loss: 77.8696, var_loss: 0.0034, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 23 - Valid Loss: 78.6770 Sub-losses: recon_loss: 78.6745, var_loss: 0.0025, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 24 - Train Loss: 78.2095 Sub-losses: recon_loss: 78.2044, var_loss: 0.0052, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 24 - Valid Loss: 78.8305 Sub-losses: recon_loss: 78.8275, var_loss: 0.0030, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 25 - Train Loss: 77.4118 Sub-losses: recon_loss: 77.4063, var_loss: 0.0055, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 25 - Valid Loss: 78.4000 Sub-losses: recon_loss: 78.3960, var_loss: 0.0041, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 26 - Train Loss: 76.8581 Sub-losses: recon_loss: 76.8507, var_loss: 0.0074, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 26 - Valid Loss: 76.6204 Sub-losses: recon_loss: 76.6154, var_loss: 0.0050, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 27 - Train Loss: 76.4866 Sub-losses: recon_loss: 76.4779, var_loss: 0.0087, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 27 - Valid Loss: 75.6482 Sub-losses: recon_loss: 75.6418, var_loss: 0.0063, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 28 - Train Loss: 76.3524 Sub-losses: recon_loss: 76.3406, var_loss: 0.0118, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 28 - Valid Loss: 75.9029 Sub-losses: recon_loss: 75.8949, var_loss: 0.0080, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 29 - Train Loss: 75.0863 Sub-losses: recon_loss: 75.0727, var_loss: 0.0135, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 29 - Valid Loss: 74.8990 Sub-losses: recon_loss: 74.8881, var_loss: 0.0109, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 30 - Train Loss: 75.2333 Sub-losses: recon_loss: 75.2144, var_loss: 0.0189, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 30 - Valid Loss: 73.6780 Sub-losses: recon_loss: 73.6630, var_loss: 0.0150, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 31 - Train Loss: 74.4064 Sub-losses: recon_loss: 74.3824, var_loss: 0.0240, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 31 - Valid Loss: 73.7967 Sub-losses: recon_loss: 73.7785, var_loss: 0.0181, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 32 - Train Loss: 74.0161 Sub-losses: recon_loss: 73.9872, var_loss: 0.0289, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 32 - Valid Loss: 73.2352 Sub-losses: recon_loss: 73.2113, var_loss: 0.0239, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 33 - Train Loss: 73.7499 Sub-losses: recon_loss: 73.7115, var_loss: 0.0383, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 33 - Valid Loss: 72.9972 Sub-losses: recon_loss: 72.9679, var_loss: 0.0292, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 34 - Train Loss: 73.4212 Sub-losses: recon_loss: 73.3720, var_loss: 0.0492, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 34 - Valid Loss: 72.4610 Sub-losses: recon_loss: 72.4241, var_loss: 0.0369, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 35 - Train Loss: 72.9689 Sub-losses: recon_loss: 72.9056, var_loss: 0.0633, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 35 - Valid Loss: 73.2469 Sub-losses: recon_loss: 73.2025, var_loss: 0.0444, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 36 - Train Loss: 72.0242 Sub-losses: recon_loss: 71.9482, var_loss: 0.0760, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 36 - Valid Loss: 71.9919 Sub-losses: recon_loss: 71.9324, var_loss: 0.0595, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 37 - Train Loss: 71.0621 Sub-losses: recon_loss: 70.9644, var_loss: 0.0977, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 37 - Valid Loss: 70.2478 Sub-losses: recon_loss: 70.1722, var_loss: 0.0755, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 38 - Train Loss: 71.6860 Sub-losses: recon_loss: 71.5677, var_loss: 0.1183, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 38 - Valid Loss: 71.1941 Sub-losses: recon_loss: 71.1028, var_loss: 0.0913, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 39 - Train Loss: 71.6087 Sub-losses: recon_loss: 71.4593, var_loss: 0.1494, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 39 - Valid Loss: 70.7408 Sub-losses: recon_loss: 70.6281, var_loss: 0.1127, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 40 - Train Loss: 70.3288 Sub-losses: recon_loss: 70.1436, var_loss: 0.1852, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 40 - Valid Loss: 69.9096 Sub-losses: recon_loss: 69.7680, var_loss: 0.1416, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 41 - Train Loss: 69.6269 Sub-losses: recon_loss: 69.4103, var_loss: 0.2166, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 41 - Valid Loss: 68.6354 Sub-losses: recon_loss: 68.4579, var_loss: 0.1775, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 42 - Train Loss: 70.6092 Sub-losses: recon_loss: 70.3405, var_loss: 0.2686, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 42 - Valid Loss: 69.1526 Sub-losses: recon_loss: 68.9319, var_loss: 0.2208, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 43 - Train Loss: 69.7526 Sub-losses: recon_loss: 69.4215, var_loss: 0.3311, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 43 - Valid Loss: 68.4814 Sub-losses: recon_loss: 68.1973, var_loss: 0.2841, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 44 - Train Loss: 69.1467 Sub-losses: recon_loss: 68.7446, var_loss: 0.4020, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 44 - Valid Loss: 67.7067 Sub-losses: recon_loss: 67.3643, var_loss: 0.3424, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 45 - Train Loss: 69.0849 Sub-losses: recon_loss: 68.5942, var_loss: 0.4907, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 45 - Valid Loss: 67.8653 Sub-losses: recon_loss: 67.4906, var_loss: 0.3747, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 46 - Train Loss: 69.3270 Sub-losses: recon_loss: 68.7549, var_loss: 0.5721, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 46 - Valid Loss: 67.9880 Sub-losses: recon_loss: 67.5356, var_loss: 0.4524, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 47 - Train Loss: 69.4533 Sub-losses: recon_loss: 68.7644, var_loss: 0.6888, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 47 - Valid Loss: 68.4470 Sub-losses: recon_loss: 67.9157, var_loss: 0.5313, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 48 - Train Loss: 69.4169 Sub-losses: recon_loss: 68.6327, var_loss: 0.7842, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 48 - Valid Loss: 67.3458 Sub-losses: recon_loss: 66.7096, var_loss: 0.6362, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 49 - Train Loss: 68.9063 Sub-losses: recon_loss: 68.0147, var_loss: 0.8916, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 49 - Valid Loss: 67.3330 Sub-losses: recon_loss: 66.5787, var_loss: 0.7544, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 50 - Train Loss: 68.9867 Sub-losses: recon_loss: 67.9992, var_loss: 0.9875, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 50 - Valid Loss: 67.5690 Sub-losses: recon_loss: 66.7118, var_loss: 0.8571, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 51 - Train Loss: 68.7537 Sub-losses: recon_loss: 67.6145, var_loss: 1.1392, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 51 - Valid Loss: 67.0762 Sub-losses: recon_loss: 66.1293, var_loss: 0.9469, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 52 - Train Loss: 68.0940 Sub-losses: recon_loss: 66.8203, var_loss: 1.2737, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 52 - Valid Loss: 68.0218 Sub-losses: recon_loss: 67.0302, var_loss: 0.9916, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 53 - Train Loss: 67.9385 Sub-losses: recon_loss: 66.5969, var_loss: 1.3417, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 53 - Valid Loss: 66.3616 Sub-losses: recon_loss: 65.2313, var_loss: 1.1303, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 54 - Train Loss: 68.2149 Sub-losses: recon_loss: 66.7506, var_loss: 1.4643, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 54 - Valid Loss: 66.5538 Sub-losses: recon_loss: 65.3370, var_loss: 1.2168, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 55 - Train Loss: 67.6376 Sub-losses: recon_loss: 66.0618, var_loss: 1.5758, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 55 - Valid Loss: 66.7092 Sub-losses: recon_loss: 65.4330, var_loss: 1.2762, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 56 - Train Loss: 68.9941 Sub-losses: recon_loss: 67.3948, var_loss: 1.5992, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 56 - Valid Loss: 66.2746 Sub-losses: recon_loss: 64.9227, var_loss: 1.3519, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 57 - Train Loss: 67.8172 Sub-losses: recon_loss: 66.1348, var_loss: 1.6824, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 57 - Valid Loss: 66.4237 Sub-losses: recon_loss: 64.9881, var_loss: 1.4356, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 58 - Train Loss: 67.0116 Sub-losses: recon_loss: 65.2516, var_loss: 1.7601, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 58 - Valid Loss: 65.2628 Sub-losses: recon_loss: 63.7581, var_loss: 1.5047, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 59 - Train Loss: 67.2695 Sub-losses: recon_loss: 65.5400, var_loss: 1.7295, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 59 - Valid Loss: 65.7285 Sub-losses: recon_loss: 64.2454, var_loss: 1.4831, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 60 - Train Loss: 67.4511 Sub-losses: recon_loss: 65.7169, var_loss: 1.7342, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 60 - Valid Loss: 64.8766 Sub-losses: recon_loss: 63.4027, var_loss: 1.4738, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 61 - Train Loss: 67.5468 Sub-losses: recon_loss: 65.7466, var_loss: 1.8002, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 61 - Valid Loss: 65.3135 Sub-losses: recon_loss: 63.8023, var_loss: 1.5112, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 62 - Train Loss: 67.2672 Sub-losses: recon_loss: 65.4704, var_loss: 1.7968, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 62 - Valid Loss: 64.6683 Sub-losses: recon_loss: 63.0695, var_loss: 1.5988, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 63 - Train Loss: 66.9896 Sub-losses: recon_loss: 65.1159, var_loss: 1.8737, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 63 - Valid Loss: 65.2256 Sub-losses: recon_loss: 63.6660, var_loss: 1.5596, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 64 - Train Loss: 67.8966 Sub-losses: recon_loss: 66.0661, var_loss: 1.8305, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 64 - Valid Loss: 64.4551 Sub-losses: recon_loss: 62.8166, var_loss: 1.6385, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 65 - Train Loss: 66.9414 Sub-losses: recon_loss: 65.0626, var_loss: 1.8789, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 65 - Valid Loss: 64.9397 Sub-losses: recon_loss: 63.3475, var_loss: 1.5922, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 66 - Train Loss: 67.1576 Sub-losses: recon_loss: 65.3460, var_loss: 1.8116, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 66 - Valid Loss: 64.9529 Sub-losses: recon_loss: 63.3179, var_loss: 1.6351, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 67 - Train Loss: 66.3333 Sub-losses: recon_loss: 64.5148, var_loss: 1.8185, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 67 - Valid Loss: 63.1336 Sub-losses: recon_loss: 61.5149, var_loss: 1.6187, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 68 - Train Loss: 67.2023 Sub-losses: recon_loss: 65.3686, var_loss: 1.8336, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 68 - Valid Loss: 62.7676 Sub-losses: recon_loss: 61.1863, var_loss: 1.5813, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 69 - Train Loss: 66.9984 Sub-losses: recon_loss: 65.1741, var_loss: 1.8243, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 69 - Valid Loss: 63.6464 Sub-losses: recon_loss: 62.0102, var_loss: 1.6362, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 70 - Train Loss: 65.7280 Sub-losses: recon_loss: 63.7886, var_loss: 1.9393, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 70 - Valid Loss: 62.9054 Sub-losses: recon_loss: 61.2777, var_loss: 1.6278, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 71 - Train Loss: 66.8908 Sub-losses: recon_loss: 65.0034, var_loss: 1.8874, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 71 - Valid Loss: 62.7513 Sub-losses: recon_loss: 61.1290, var_loss: 1.6223, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 72 - Train Loss: 66.2318 Sub-losses: recon_loss: 64.3862, var_loss: 1.8456, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 72 - Valid Loss: 63.2070 Sub-losses: recon_loss: 61.5386, var_loss: 1.6684, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 73 - Train Loss: 66.0335 Sub-losses: recon_loss: 64.1350, var_loss: 1.8985, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 73 - Valid Loss: 62.3002 Sub-losses: recon_loss: 60.7131, var_loss: 1.5871, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 74 - Train Loss: 66.1059 Sub-losses: recon_loss: 64.2238, var_loss: 1.8821, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 74 - Valid Loss: 61.3905 Sub-losses: recon_loss: 59.7137, var_loss: 1.6768, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 75 - Train Loss: 65.8689 Sub-losses: recon_loss: 63.9560, var_loss: 1.9129, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 75 - Valid Loss: 62.7231 Sub-losses: recon_loss: 61.0619, var_loss: 1.6612, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 76 - Train Loss: 65.4043 Sub-losses: recon_loss: 63.5866, var_loss: 1.8176, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 76 - Valid Loss: 62.3085 Sub-losses: recon_loss: 60.6495, var_loss: 1.6590, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 77 - Train Loss: 64.8734 Sub-losses: recon_loss: 62.9981, var_loss: 1.8753, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 77 - Valid Loss: 62.6655 Sub-losses: recon_loss: 60.9721, var_loss: 1.6934, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 78 - Train Loss: 65.5895 Sub-losses: recon_loss: 63.7487, var_loss: 1.8408, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 78 - Valid Loss: 62.2469 Sub-losses: recon_loss: 60.5746, var_loss: 1.6723, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 79 - Train Loss: 65.8088 Sub-losses: recon_loss: 63.9639, var_loss: 1.8448, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 79 - Valid Loss: 61.2911 Sub-losses: recon_loss: 59.5882, var_loss: 1.7029, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 80 - Train Loss: 65.4850 Sub-losses: recon_loss: 63.6288, var_loss: 1.8562, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 80 - Valid Loss: 61.3704 Sub-losses: recon_loss: 59.6188, var_loss: 1.7516, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 81 - Train Loss: 65.0258 Sub-losses: recon_loss: 63.1641, var_loss: 1.8617, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 81 - Valid Loss: 60.9812 Sub-losses: recon_loss: 59.2806, var_loss: 1.7006, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 82 - Train Loss: 65.6162 Sub-losses: recon_loss: 63.6786, var_loss: 1.9376, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 82 - Valid Loss: 61.5507 Sub-losses: recon_loss: 59.9192, var_loss: 1.6315, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 83 - Train Loss: 64.3936 Sub-losses: recon_loss: 62.4729, var_loss: 1.9207, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 83 - Valid Loss: 61.4858 Sub-losses: recon_loss: 59.8787, var_loss: 1.6071, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 84 - Train Loss: 65.0223 Sub-losses: recon_loss: 63.1470, var_loss: 1.8753, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 84 - Valid Loss: 61.6351 Sub-losses: recon_loss: 59.9934, var_loss: 1.6417, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 85 - Train Loss: 64.8710 Sub-losses: recon_loss: 63.0000, var_loss: 1.8710, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 85 - Valid Loss: 60.4532 Sub-losses: recon_loss: 58.7682, var_loss: 1.6849, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 86 - Train Loss: 65.1813 Sub-losses: recon_loss: 63.3080, var_loss: 1.8733, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 86 - Valid Loss: 59.9558 Sub-losses: recon_loss: 58.2224, var_loss: 1.7334, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 87 - Train Loss: 64.3401 Sub-losses: recon_loss: 62.3805, var_loss: 1.9596, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 87 - Valid Loss: 60.1070 Sub-losses: recon_loss: 58.3111, var_loss: 1.7959, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 88 - Train Loss: 64.1191 Sub-losses: recon_loss: 62.1460, var_loss: 1.9730, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 88 - Valid Loss: 60.2929 Sub-losses: recon_loss: 58.5788, var_loss: 1.7141, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 89 - Train Loss: 65.1576 Sub-losses: recon_loss: 63.1989, var_loss: 1.9587, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 89 - Valid Loss: 60.5760 Sub-losses: recon_loss: 58.7788, var_loss: 1.7972, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 90 - Train Loss: 63.4165 Sub-losses: recon_loss: 61.4205, var_loss: 1.9960, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 90 - Valid Loss: 59.8966 Sub-losses: recon_loss: 58.1310, var_loss: 1.7655, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 91 - Train Loss: 64.4972 Sub-losses: recon_loss: 62.5106, var_loss: 1.9865, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 91 - Valid Loss: 60.0542 Sub-losses: recon_loss: 58.2384, var_loss: 1.8159, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Train Loss: 63.2402 Sub-losses: recon_loss: 61.2669, var_loss: 1.9733, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Valid Loss: 59.8329 Sub-losses: recon_loss: 57.9815, var_loss: 1.8514, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 93 - Train Loss: 63.3848 Sub-losses: recon_loss: 61.3798, var_loss: 2.0050, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 93 - Valid Loss: 59.6810 Sub-losses: recon_loss: 57.8938, var_loss: 1.7872, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Train Loss: 64.3425 Sub-losses: recon_loss: 62.3974, var_loss: 1.9452, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Valid Loss: 59.3407 Sub-losses: recon_loss: 57.4957, var_loss: 1.8450, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Train Loss: 64.0368 Sub-losses: recon_loss: 62.0326, var_loss: 2.0041, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Valid Loss: 58.7256 Sub-losses: recon_loss: 56.8609, var_loss: 1.8647, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 96 - Train Loss: 64.4129 Sub-losses: recon_loss: 62.3998, var_loss: 2.0132, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 96 - Valid Loss: 58.6279 Sub-losses: recon_loss: 56.7908, var_loss: 1.8371, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Train Loss: 62.4973 Sub-losses: recon_loss: 60.5013, var_loss: 1.9960, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Valid Loss: 59.7730 Sub-losses: recon_loss: 57.9385, var_loss: 1.8345, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Train Loss: 63.3329 Sub-losses: recon_loss: 61.3302, var_loss: 2.0028, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Valid Loss: 59.8923 Sub-losses: recon_loss: 58.1080, var_loss: 1.7842, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Train Loss: 62.4985 Sub-losses: recon_loss: 60.4059, var_loss: 2.0925, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Valid Loss: 58.7958 Sub-losses: recon_loss: 56.9905, var_loss: 1.8053, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Train Loss: 63.3659 Sub-losses: recon_loss: 61.3732, var_loss: 1.9928, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Valid Loss: 58.3032 Sub-losses: recon_loss: 56.5318, var_loss: 1.7714, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Found 350 common samples for the stacked autoencoder. Found 50 common samples for the stacked autoencoder. finished training each modality model Epoch 1 - Train Loss: 122.1290 Sub-losses: recon_loss: 122.1290, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 1 - Valid Loss: 110.3874 Sub-losses: recon_loss: 110.3874, var_loss: 0.0000, anneal_factor: 0.0000, effective_beta_factor: 0.0000 Epoch 2 - Train Loss: 118.3054 Sub-losses: recon_loss: 118.3053, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 2 - Valid Loss: 108.2109 Sub-losses: recon_loss: 108.2109, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Train Loss: 113.3633 Sub-losses: recon_loss: 113.3633, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 3 - Valid Loss: 106.3902 Sub-losses: recon_loss: 106.3902, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Train Loss: 110.6027 Sub-losses: recon_loss: 110.6027, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 4 - Valid Loss: 103.9404 Sub-losses: recon_loss: 103.9404, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Train Loss: 107.9854 Sub-losses: recon_loss: 107.9854, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 5 - Valid Loss: 101.9015 Sub-losses: recon_loss: 101.9015, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Train Loss: 105.3416 Sub-losses: recon_loss: 105.3415, var_loss: 0.0001, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 6 - Valid Loss: 101.0094 Sub-losses: recon_loss: 101.0094, var_loss: 0.0000, anneal_factor: 0.0001, effective_beta_factor: 0.0000 Epoch 7 - Train Loss: 102.9300 Sub-losses: recon_loss: 102.9300, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 7 - Valid Loss: 98.2106 Sub-losses: recon_loss: 98.2106, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Train Loss: 100.2904 Sub-losses: recon_loss: 100.2903, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 8 - Valid Loss: 96.5362 Sub-losses: recon_loss: 96.5362, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Train Loss: 98.4548 Sub-losses: recon_loss: 98.4547, var_loss: 0.0001, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 9 - Valid Loss: 94.9954 Sub-losses: recon_loss: 94.9953, var_loss: 0.0000, anneal_factor: 0.0002, effective_beta_factor: 0.0000 Epoch 10 - Train Loss: 97.4701 Sub-losses: recon_loss: 97.4700, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 10 - Valid Loss: 93.0627 Sub-losses: recon_loss: 93.0626, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Train Loss: 96.3597 Sub-losses: recon_loss: 96.3596, var_loss: 0.0002, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 11 - Valid Loss: 92.4208 Sub-losses: recon_loss: 92.4208, var_loss: 0.0001, anneal_factor: 0.0003, effective_beta_factor: 0.0000 Epoch 12 - Train Loss: 95.0307 Sub-losses: recon_loss: 95.0305, var_loss: 0.0002, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 12 - Valid Loss: 89.6648 Sub-losses: recon_loss: 89.6647, var_loss: 0.0001, anneal_factor: 0.0004, effective_beta_factor: 0.0000 Epoch 13 - Train Loss: 93.7278 Sub-losses: recon_loss: 93.7276, var_loss: 0.0002, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 13 - Valid Loss: 89.3395 Sub-losses: recon_loss: 89.3393, var_loss: 0.0001, anneal_factor: 0.0005, effective_beta_factor: 0.0001 Epoch 14 - Train Loss: 93.6053 Sub-losses: recon_loss: 93.6050, var_loss: 0.0003, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 14 - Valid Loss: 87.8521 Sub-losses: recon_loss: 87.8520, var_loss: 0.0001, anneal_factor: 0.0006, effective_beta_factor: 0.0001 Epoch 15 - Train Loss: 92.2374 Sub-losses: recon_loss: 92.2370, var_loss: 0.0004, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 15 - Valid Loss: 85.9753 Sub-losses: recon_loss: 85.9751, var_loss: 0.0002, anneal_factor: 0.0007, effective_beta_factor: 0.0001 Epoch 16 - Train Loss: 91.1066 Sub-losses: recon_loss: 91.1062, var_loss: 0.0005, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 16 - Valid Loss: 85.6668 Sub-losses: recon_loss: 85.6666, var_loss: 0.0002, anneal_factor: 0.0009, effective_beta_factor: 0.0001 Epoch 17 - Train Loss: 89.8455 Sub-losses: recon_loss: 89.8449, var_loss: 0.0006, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 17 - Valid Loss: 84.6326 Sub-losses: recon_loss: 84.6323, var_loss: 0.0003, anneal_factor: 0.0011, effective_beta_factor: 0.0001 Epoch 18 - Train Loss: 89.3785 Sub-losses: recon_loss: 89.3777, var_loss: 0.0007, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 18 - Valid Loss: 84.5200 Sub-losses: recon_loss: 84.5196, var_loss: 0.0004, anneal_factor: 0.0014, effective_beta_factor: 0.0001 Epoch 19 - Train Loss: 88.6056 Sub-losses: recon_loss: 88.6046, var_loss: 0.0010, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 19 - Valid Loss: 83.0993 Sub-losses: recon_loss: 83.0988, var_loss: 0.0005, anneal_factor: 0.0017, effective_beta_factor: 0.0002 Epoch 20 - Train Loss: 87.8259 Sub-losses: recon_loss: 87.8247, var_loss: 0.0012, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 20 - Valid Loss: 81.6810 Sub-losses: recon_loss: 81.6803, var_loss: 0.0006, anneal_factor: 0.0020, effective_beta_factor: 0.0002 Epoch 21 - Train Loss: 86.5908 Sub-losses: recon_loss: 86.5893, var_loss: 0.0015, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 21 - Valid Loss: 82.3074 Sub-losses: recon_loss: 82.3065, var_loss: 0.0009, anneal_factor: 0.0025, effective_beta_factor: 0.0002 Epoch 22 - Train Loss: 86.3044 Sub-losses: recon_loss: 86.3026, var_loss: 0.0018, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 22 - Valid Loss: 79.6072 Sub-losses: recon_loss: 79.6061, var_loss: 0.0011, anneal_factor: 0.0030, effective_beta_factor: 0.0003 Epoch 23 - Train Loss: 85.8850 Sub-losses: recon_loss: 85.8826, var_loss: 0.0024, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 23 - Valid Loss: 80.5318 Sub-losses: recon_loss: 80.5303, var_loss: 0.0014, anneal_factor: 0.0037, effective_beta_factor: 0.0004 Epoch 24 - Train Loss: 84.5413 Sub-losses: recon_loss: 84.5383, var_loss: 0.0030, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 24 - Valid Loss: 78.9876 Sub-losses: recon_loss: 78.9858, var_loss: 0.0018, anneal_factor: 0.0045, effective_beta_factor: 0.0004 Epoch 25 - Train Loss: 85.4577 Sub-losses: recon_loss: 85.4538, var_loss: 0.0038, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 25 - Valid Loss: 77.8169 Sub-losses: recon_loss: 77.8145, var_loss: 0.0023, anneal_factor: 0.0055, effective_beta_factor: 0.0005 Epoch 26 - Train Loss: 83.6593 Sub-losses: recon_loss: 83.6545, var_loss: 0.0048, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 26 - Valid Loss: 78.8694 Sub-losses: recon_loss: 78.8664, var_loss: 0.0030, anneal_factor: 0.0067, effective_beta_factor: 0.0007 Epoch 27 - Train Loss: 83.6343 Sub-losses: recon_loss: 83.6280, var_loss: 0.0063, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 27 - Valid Loss: 75.8494 Sub-losses: recon_loss: 75.8453, var_loss: 0.0041, anneal_factor: 0.0082, effective_beta_factor: 0.0008 Epoch 28 - Train Loss: 83.2374 Sub-losses: recon_loss: 83.2296, var_loss: 0.0077, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 28 - Valid Loss: 74.3409 Sub-losses: recon_loss: 74.3357, var_loss: 0.0052, anneal_factor: 0.0100, effective_beta_factor: 0.0010 Epoch 29 - Train Loss: 83.2226 Sub-losses: recon_loss: 83.2131, var_loss: 0.0095, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 29 - Valid Loss: 75.1699 Sub-losses: recon_loss: 75.1631, var_loss: 0.0068, anneal_factor: 0.0121, effective_beta_factor: 0.0012 Epoch 30 - Train Loss: 81.6765 Sub-losses: recon_loss: 81.6640, var_loss: 0.0126, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 30 - Valid Loss: 73.4028 Sub-losses: recon_loss: 73.3937, var_loss: 0.0092, anneal_factor: 0.0148, effective_beta_factor: 0.0015 Epoch 31 - Train Loss: 81.4690 Sub-losses: recon_loss: 81.4527, var_loss: 0.0163, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 31 - Valid Loss: 75.5304 Sub-losses: recon_loss: 75.5194, var_loss: 0.0109, anneal_factor: 0.0180, effective_beta_factor: 0.0018 Epoch 32 - Train Loss: 81.2880 Sub-losses: recon_loss: 81.2682, var_loss: 0.0198, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 32 - Valid Loss: 74.1153 Sub-losses: recon_loss: 74.1019, var_loss: 0.0134, anneal_factor: 0.0219, effective_beta_factor: 0.0022 Epoch 33 - Train Loss: 81.2565 Sub-losses: recon_loss: 81.2313, var_loss: 0.0252, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 33 - Valid Loss: 72.4740 Sub-losses: recon_loss: 72.4575, var_loss: 0.0165, anneal_factor: 0.0266, effective_beta_factor: 0.0027 Epoch 34 - Train Loss: 79.6420 Sub-losses: recon_loss: 79.6104, var_loss: 0.0317, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 34 - Valid Loss: 73.6142 Sub-losses: recon_loss: 73.5920, var_loss: 0.0222, anneal_factor: 0.0323, effective_beta_factor: 0.0032 Epoch 35 - Train Loss: 79.9643 Sub-losses: recon_loss: 79.9254, var_loss: 0.0388, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 35 - Valid Loss: 71.2553 Sub-losses: recon_loss: 71.2249, var_loss: 0.0304, anneal_factor: 0.0392, effective_beta_factor: 0.0039 Epoch 36 - Train Loss: 79.3759 Sub-losses: recon_loss: 79.3276, var_loss: 0.0483, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 36 - Valid Loss: 70.7066 Sub-losses: recon_loss: 70.6707, var_loss: 0.0359, anneal_factor: 0.0474, effective_beta_factor: 0.0047 Epoch 37 - Train Loss: 77.2487 Sub-losses: recon_loss: 77.1828, var_loss: 0.0660, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 37 - Valid Loss: 71.3753 Sub-losses: recon_loss: 71.3316, var_loss: 0.0438, anneal_factor: 0.0573, effective_beta_factor: 0.0057 Epoch 38 - Train Loss: 78.9325 Sub-losses: recon_loss: 78.8528, var_loss: 0.0796, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 38 - Valid Loss: 71.8148 Sub-losses: recon_loss: 71.7564, var_loss: 0.0584, anneal_factor: 0.0691, effective_beta_factor: 0.0069 Epoch 39 - Train Loss: 76.5608 Sub-losses: recon_loss: 76.4632, var_loss: 0.0976, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 39 - Valid Loss: 70.9236 Sub-losses: recon_loss: 70.8531, var_loss: 0.0705, anneal_factor: 0.0832, effective_beta_factor: 0.0083 Epoch 40 - Train Loss: 75.9799 Sub-losses: recon_loss: 75.8508, var_loss: 0.1290, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 40 - Valid Loss: 69.3154 Sub-losses: recon_loss: 69.2201, var_loss: 0.0953, anneal_factor: 0.0998, effective_beta_factor: 0.0100 Epoch 41 - Train Loss: 76.0444 Sub-losses: recon_loss: 75.8954, var_loss: 0.1490, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 41 - Valid Loss: 68.5774 Sub-losses: recon_loss: 68.4591, var_loss: 0.1183, anneal_factor: 0.1192, effective_beta_factor: 0.0119 Epoch 42 - Train Loss: 76.0765 Sub-losses: recon_loss: 75.8889, var_loss: 0.1875, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 42 - Valid Loss: 68.4036 Sub-losses: recon_loss: 68.2591, var_loss: 0.1444, anneal_factor: 0.1419, effective_beta_factor: 0.0142 Epoch 43 - Train Loss: 75.8277 Sub-losses: recon_loss: 75.6040, var_loss: 0.2237, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 43 - Valid Loss: 66.7183 Sub-losses: recon_loss: 66.5367, var_loss: 0.1816, anneal_factor: 0.1680, effective_beta_factor: 0.0168 Epoch 44 - Train Loss: 75.7542 Sub-losses: recon_loss: 75.4737, var_loss: 0.2805, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 44 - Valid Loss: 66.6314 Sub-losses: recon_loss: 66.4093, var_loss: 0.2221, anneal_factor: 0.1978, effective_beta_factor: 0.0198 Epoch 45 - Train Loss: 74.3553 Sub-losses: recon_loss: 74.0193, var_loss: 0.3360, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 45 - Valid Loss: 66.4539 Sub-losses: recon_loss: 66.1847, var_loss: 0.2691, anneal_factor: 0.2315, effective_beta_factor: 0.0231 Epoch 46 - Train Loss: 74.5574 Sub-losses: recon_loss: 74.1640, var_loss: 0.3934, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 46 - Valid Loss: 65.8863 Sub-losses: recon_loss: 65.5564, var_loss: 0.3300, anneal_factor: 0.2689, effective_beta_factor: 0.0269 Epoch 47 - Train Loss: 73.6675 Sub-losses: recon_loss: 73.1928, var_loss: 0.4747, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 47 - Valid Loss: 65.4339 Sub-losses: recon_loss: 65.0381, var_loss: 0.3958, anneal_factor: 0.3100, effective_beta_factor: 0.0310 Epoch 48 - Train Loss: 73.6270 Sub-losses: recon_loss: 73.0704, var_loss: 0.5566, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 48 - Valid Loss: 65.5841 Sub-losses: recon_loss: 65.1419, var_loss: 0.4421, anneal_factor: 0.3543, effective_beta_factor: 0.0354 Epoch 49 - Train Loss: 73.2195 Sub-losses: recon_loss: 72.5825, var_loss: 0.6370, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 49 - Valid Loss: 63.9188 Sub-losses: recon_loss: 63.3739, var_loss: 0.5450, anneal_factor: 0.4013, effective_beta_factor: 0.0401 Epoch 50 - Train Loss: 73.5603 Sub-losses: recon_loss: 72.8153, var_loss: 0.7450, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 50 - Valid Loss: 63.3327 Sub-losses: recon_loss: 62.6940, var_loss: 0.6387, anneal_factor: 0.4502, effective_beta_factor: 0.0450 Epoch 51 - Train Loss: 72.5836 Sub-losses: recon_loss: 71.7664, var_loss: 0.8171, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 51 - Valid Loss: 63.9558 Sub-losses: recon_loss: 63.2330, var_loss: 0.7228, anneal_factor: 0.5000, effective_beta_factor: 0.0500 Epoch 52 - Train Loss: 72.7734 Sub-losses: recon_loss: 71.8751, var_loss: 0.8984, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 52 - Valid Loss: 63.8811 Sub-losses: recon_loss: 63.0843, var_loss: 0.7968, anneal_factor: 0.5498, effective_beta_factor: 0.0550 Epoch 53 - Train Loss: 71.7590 Sub-losses: recon_loss: 70.7288, var_loss: 1.0302, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 53 - Valid Loss: 63.1551 Sub-losses: recon_loss: 62.3122, var_loss: 0.8430, anneal_factor: 0.5987, effective_beta_factor: 0.0599 Epoch 54 - Train Loss: 72.4263 Sub-losses: recon_loss: 71.3172, var_loss: 1.1091, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 54 - Valid Loss: 63.0428 Sub-losses: recon_loss: 62.1169, var_loss: 0.9259, anneal_factor: 0.6457, effective_beta_factor: 0.0646 Epoch 55 - Train Loss: 72.8788 Sub-losses: recon_loss: 71.6960, var_loss: 1.1828, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 55 - Valid Loss: 61.7804 Sub-losses: recon_loss: 60.7224, var_loss: 1.0580, anneal_factor: 0.6900, effective_beta_factor: 0.0690 Epoch 56 - Train Loss: 72.9134 Sub-losses: recon_loss: 71.6310, var_loss: 1.2825, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 56 - Valid Loss: 62.6996 Sub-losses: recon_loss: 61.6122, var_loss: 1.0875, anneal_factor: 0.7311, effective_beta_factor: 0.0731 Epoch 57 - Train Loss: 71.5684 Sub-losses: recon_loss: 70.2335, var_loss: 1.3349, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 57 - Valid Loss: 62.9758 Sub-losses: recon_loss: 61.8088, var_loss: 1.1670, anneal_factor: 0.7685, effective_beta_factor: 0.0769 Epoch 58 - Train Loss: 71.3943 Sub-losses: recon_loss: 70.0091, var_loss: 1.3852, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 58 - Valid Loss: 62.1583 Sub-losses: recon_loss: 60.9374, var_loss: 1.2209, anneal_factor: 0.8022, effective_beta_factor: 0.0802 Epoch 59 - Train Loss: 72.1014 Sub-losses: recon_loss: 70.6966, var_loss: 1.4048, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 59 - Valid Loss: 61.5845 Sub-losses: recon_loss: 60.3199, var_loss: 1.2646, anneal_factor: 0.8320, effective_beta_factor: 0.0832 Epoch 60 - Train Loss: 70.5105 Sub-losses: recon_loss: 69.0225, var_loss: 1.4880, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 60 - Valid Loss: 62.2785 Sub-losses: recon_loss: 60.9609, var_loss: 1.3176, anneal_factor: 0.8581, effective_beta_factor: 0.0858 Epoch 61 - Train Loss: 70.7415 Sub-losses: recon_loss: 69.2493, var_loss: 1.4922, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 61 - Valid Loss: 61.7770 Sub-losses: recon_loss: 60.3815, var_loss: 1.3955, anneal_factor: 0.8808, effective_beta_factor: 0.0881 Epoch 62 - Train Loss: 70.9902 Sub-losses: recon_loss: 69.3991, var_loss: 1.5912, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 62 - Valid Loss: 61.3474 Sub-losses: recon_loss: 59.9118, var_loss: 1.4355, anneal_factor: 0.9002, effective_beta_factor: 0.0900 Epoch 63 - Train Loss: 70.0570 Sub-losses: recon_loss: 68.4298, var_loss: 1.6272, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 63 - Valid Loss: 60.8208 Sub-losses: recon_loss: 59.3178, var_loss: 1.5030, anneal_factor: 0.9168, effective_beta_factor: 0.0917 Epoch 64 - Train Loss: 71.0984 Sub-losses: recon_loss: 69.4148, var_loss: 1.6836, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 64 - Valid Loss: 61.4759 Sub-losses: recon_loss: 59.9698, var_loss: 1.5061, anneal_factor: 0.9309, effective_beta_factor: 0.0931 Epoch 65 - Train Loss: 70.4108 Sub-losses: recon_loss: 68.7876, var_loss: 1.6232, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 65 - Valid Loss: 60.6531 Sub-losses: recon_loss: 59.1675, var_loss: 1.4856, anneal_factor: 0.9427, effective_beta_factor: 0.0943 Epoch 66 - Train Loss: 70.1082 Sub-losses: recon_loss: 68.3941, var_loss: 1.7141, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 66 - Valid Loss: 60.4393 Sub-losses: recon_loss: 58.9210, var_loss: 1.5184, anneal_factor: 0.9526, effective_beta_factor: 0.0953 Epoch 67 - Train Loss: 70.4872 Sub-losses: recon_loss: 68.8145, var_loss: 1.6728, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 67 - Valid Loss: 60.7125 Sub-losses: recon_loss: 59.1961, var_loss: 1.5164, anneal_factor: 0.9608, effective_beta_factor: 0.0961 Epoch 68 - Train Loss: 69.5528 Sub-losses: recon_loss: 67.7981, var_loss: 1.7547, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 68 - Valid Loss: 60.3154 Sub-losses: recon_loss: 58.7556, var_loss: 1.5598, anneal_factor: 0.9677, effective_beta_factor: 0.0968 Epoch 69 - Train Loss: 70.4036 Sub-losses: recon_loss: 68.6664, var_loss: 1.7372, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 69 - Valid Loss: 61.0808 Sub-losses: recon_loss: 59.4728, var_loss: 1.6080, anneal_factor: 0.9734, effective_beta_factor: 0.0973 Epoch 70 - Train Loss: 69.4210 Sub-losses: recon_loss: 67.6237, var_loss: 1.7974, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 70 - Valid Loss: 60.9774 Sub-losses: recon_loss: 59.4180, var_loss: 1.5594, anneal_factor: 0.9781, effective_beta_factor: 0.0978 Epoch 71 - Train Loss: 69.4865 Sub-losses: recon_loss: 67.7163, var_loss: 1.7702, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 71 - Valid Loss: 58.4749 Sub-losses: recon_loss: 56.8972, var_loss: 1.5776, anneal_factor: 0.9820, effective_beta_factor: 0.0982 Epoch 72 - Train Loss: 69.7624 Sub-losses: recon_loss: 68.0256, var_loss: 1.7369, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 72 - Valid Loss: 59.0071 Sub-losses: recon_loss: 57.4124, var_loss: 1.5947, anneal_factor: 0.9852, effective_beta_factor: 0.0985 Epoch 73 - Train Loss: 70.3216 Sub-losses: recon_loss: 68.5329, var_loss: 1.7887, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 73 - Valid Loss: 60.0950 Sub-losses: recon_loss: 58.4760, var_loss: 1.6190, anneal_factor: 0.9879, effective_beta_factor: 0.0988 Epoch 74 - Train Loss: 69.5782 Sub-losses: recon_loss: 67.8021, var_loss: 1.7761, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 74 - Valid Loss: 59.7582 Sub-losses: recon_loss: 58.0940, var_loss: 1.6643, anneal_factor: 0.9900, effective_beta_factor: 0.0990 Epoch 75 - Train Loss: 68.5203 Sub-losses: recon_loss: 66.7053, var_loss: 1.8150, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 75 - Valid Loss: 58.4155 Sub-losses: recon_loss: 56.8041, var_loss: 1.6114, anneal_factor: 0.9918, effective_beta_factor: 0.0992 Epoch 76 - Train Loss: 68.8883 Sub-losses: recon_loss: 67.0606, var_loss: 1.8277, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 76 - Valid Loss: 58.6514 Sub-losses: recon_loss: 57.1069, var_loss: 1.5445, anneal_factor: 0.9933, effective_beta_factor: 0.0993 Epoch 77 - Train Loss: 69.5132 Sub-losses: recon_loss: 67.7202, var_loss: 1.7930, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 77 - Valid Loss: 59.2427 Sub-losses: recon_loss: 57.6947, var_loss: 1.5480, anneal_factor: 0.9945, effective_beta_factor: 0.0995 Epoch 78 - Train Loss: 69.4810 Sub-losses: recon_loss: 67.6612, var_loss: 1.8198, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 78 - Valid Loss: 59.4732 Sub-losses: recon_loss: 57.8426, var_loss: 1.6306, anneal_factor: 0.9955, effective_beta_factor: 0.0996 Epoch 79 - Train Loss: 68.4903 Sub-losses: recon_loss: 66.6703, var_loss: 1.8200, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 79 - Valid Loss: 59.0616 Sub-losses: recon_loss: 57.4632, var_loss: 1.5984, anneal_factor: 0.9963, effective_beta_factor: 0.0996 Epoch 80 - Train Loss: 69.4440 Sub-losses: recon_loss: 67.5845, var_loss: 1.8595, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 80 - Valid Loss: 59.5720 Sub-losses: recon_loss: 57.9472, var_loss: 1.6248, anneal_factor: 0.9970, effective_beta_factor: 0.0997 Epoch 81 - Train Loss: 68.1564 Sub-losses: recon_loss: 66.3010, var_loss: 1.8553, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 81 - Valid Loss: 58.7668 Sub-losses: recon_loss: 57.1413, var_loss: 1.6256, anneal_factor: 0.9975, effective_beta_factor: 0.0998 Epoch 82 - Train Loss: 68.0079 Sub-losses: recon_loss: 66.2047, var_loss: 1.8032, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 82 - Valid Loss: 58.2523 Sub-losses: recon_loss: 56.5918, var_loss: 1.6605, anneal_factor: 0.9980, effective_beta_factor: 0.0998 Epoch 83 - Train Loss: 67.9784 Sub-losses: recon_loss: 66.1629, var_loss: 1.8155, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 83 - Valid Loss: 59.2659 Sub-losses: recon_loss: 57.6061, var_loss: 1.6598, anneal_factor: 0.9983, effective_beta_factor: 0.0998 Epoch 84 - Train Loss: 68.0539 Sub-losses: recon_loss: 66.2223, var_loss: 1.8316, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 84 - Valid Loss: 58.9200 Sub-losses: recon_loss: 57.2969, var_loss: 1.6231, anneal_factor: 0.9986, effective_beta_factor: 0.0999 Epoch 85 - Train Loss: 67.0794 Sub-losses: recon_loss: 65.2434, var_loss: 1.8360, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 85 - Valid Loss: 58.2363 Sub-losses: recon_loss: 56.6434, var_loss: 1.5929, anneal_factor: 0.9989, effective_beta_factor: 0.0999 Epoch 86 - Train Loss: 67.2396 Sub-losses: recon_loss: 65.4411, var_loss: 1.7985, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 86 - Valid Loss: 58.2252 Sub-losses: recon_loss: 56.6491, var_loss: 1.5761, anneal_factor: 0.9991, effective_beta_factor: 0.0999 Epoch 87 - Train Loss: 68.0519 Sub-losses: recon_loss: 66.2663, var_loss: 1.7857, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 87 - Valid Loss: 57.8245 Sub-losses: recon_loss: 56.1703, var_loss: 1.6542, anneal_factor: 0.9993, effective_beta_factor: 0.0999 Epoch 88 - Train Loss: 67.3886 Sub-losses: recon_loss: 65.5250, var_loss: 1.8636, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 88 - Valid Loss: 57.9777 Sub-losses: recon_loss: 56.3866, var_loss: 1.5910, anneal_factor: 0.9994, effective_beta_factor: 0.0999 Epoch 89 - Train Loss: 68.1131 Sub-losses: recon_loss: 66.2902, var_loss: 1.8229, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 89 - Valid Loss: 57.5601 Sub-losses: recon_loss: 55.9736, var_loss: 1.5865, anneal_factor: 0.9995, effective_beta_factor: 0.0999 Epoch 90 - Train Loss: 67.1251 Sub-losses: recon_loss: 65.3469, var_loss: 1.7782, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 90 - Valid Loss: 57.8967 Sub-losses: recon_loss: 56.3121, var_loss: 1.5846, anneal_factor: 0.9996, effective_beta_factor: 0.1000 Epoch 91 - Train Loss: 67.1166 Sub-losses: recon_loss: 65.2823, var_loss: 1.8343, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 91 - Valid Loss: 57.6123 Sub-losses: recon_loss: 56.0085, var_loss: 1.6038, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Train Loss: 66.4955 Sub-losses: recon_loss: 64.6279, var_loss: 1.8677, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 92 - Valid Loss: 57.2528 Sub-losses: recon_loss: 55.6286, var_loss: 1.6242, anneal_factor: 0.9997, effective_beta_factor: 0.1000 Epoch 93 - Train Loss: 67.4577 Sub-losses: recon_loss: 65.5692, var_loss: 1.8885, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 93 - Valid Loss: 57.5054 Sub-losses: recon_loss: 55.8463, var_loss: 1.6592, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Train Loss: 67.4787 Sub-losses: recon_loss: 65.6030, var_loss: 1.8757, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 94 - Valid Loss: 57.4144 Sub-losses: recon_loss: 55.7513, var_loss: 1.6631, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Train Loss: 66.1896 Sub-losses: recon_loss: 64.2945, var_loss: 1.8951, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 95 - Valid Loss: 57.2410 Sub-losses: recon_loss: 55.5446, var_loss: 1.6964, anneal_factor: 0.9998, effective_beta_factor: 0.1000 Epoch 96 - Train Loss: 66.3494 Sub-losses: recon_loss: 64.4343, var_loss: 1.9151, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 96 - Valid Loss: 57.7297 Sub-losses: recon_loss: 56.0854, var_loss: 1.6443, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Train Loss: 67.2721 Sub-losses: recon_loss: 65.3129, var_loss: 1.9592, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 97 - Valid Loss: 57.8856 Sub-losses: recon_loss: 56.1742, var_loss: 1.7114, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Train Loss: 67.1014 Sub-losses: recon_loss: 65.1700, var_loss: 1.9313, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 98 - Valid Loss: 57.1837 Sub-losses: recon_loss: 55.5650, var_loss: 1.6187, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Train Loss: 67.0608 Sub-losses: recon_loss: 65.1855, var_loss: 1.8753, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 99 - Valid Loss: 56.9043 Sub-losses: recon_loss: 55.1829, var_loss: 1.7214, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Train Loss: 65.7403 Sub-losses: recon_loss: 63.7559, var_loss: 1.9843, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Epoch 100 - Valid Loss: 56.3750 Sub-losses: recon_loss: 54.6625, var_loss: 1.7125, anneal_factor: 0.9999, effective_beta_factor: 0.1000 Found 100 common samples for the stacked autoencoder. <autoencodix.data._numeric_dataset.NumericDataset object at 0x3218216f0> Successfully created annotated latent space object (adata_latent).
Before we perform embedding evaluation on a metadata parameter. Let's check visually if latent space shows discrimination regarding this parameter.
In [2]:
Copied!
my_target_param = "condition"
stackix.visualizer.show_latent_space(
result= result,
plot_type="Ridgeline",
param= [my_target_param]
)
my_target_param = "condition"
stackix.visualizer.show_latent_space(
result= result,
plot_type="Ridgeline",
param= [my_target_param]
)
We can see a separation of conditions along several dimension. Hence, we can expect that embeddings are useful to classify.
TODO make example stackix better
Now we can call the pipeline step stackix.evaluate()
In [3]:
Copied!
result = stackix.evaluate(params=[my_target_param])
print("Evaluation results:")
print(result.embedding_evaluation)
result = stackix.evaluate(params=[my_target_param])
print("Evaluation results:")
print(result.embedding_evaluation)
Perform ML task with feature df: Latent
Latent
Perform ML task for target parameter: condition
Evaluation results:
score_split CLINIC_PARAM metric value ML_ALG \
0 train condition roc_auc_ovo 0.718757 LogisticRegression()
1 valid condition roc_auc_ovo 0.752485 LogisticRegression()
2 test condition roc_auc_ovo 0.673119 LogisticRegression()
ML_TYPE ML_TASK ML_SUBTASK
0 classification Latent Latent
1 classification Latent Latent
2 classification Latent Latent
From the dataframe containing the results, we can see that default configuration of .evaluate() uses linear machine learning models for regression and classification tasks.
But evaluate offers many options for a detailed inspection:
- With
ml_model_classandml_model_regressionuser can provide any machine learning method if they are using a sklearn-like interface. params- Is the list of metadata parameter on which downstream tasks will be performed. If not a list, but the string "all" is provided, all parameters will be evaluated. Type of task classification vs. regression will be auto-detected. Make sure to avoid providing class information as integers or floats.- With
metric_classandmetric_regressionthe metric for evaluation is specified for classification (Default:roc_auc_ovo) or regression (Default:r2). Must be one as described in sklearn: https://scikit-learn.org/stable/modules/model_evaluation.html#scoring-string-names split_type- Specifies how data should be splitted for training and testing downstream ML models. Default isuse-splitwhich will use established split and will train on the train split. Alternatively and in case there is no train split, cross-validation can be performed by using a string like "LOOCV", "CV-5", ... "CV-10", where the number specifies the number of folds.n_downsample- To speed up evaluation on large data sets, evaluation can be performed on a random subset. Default is 10000. You may need to increase in case of extreme class-imbalance.reference_methods- For cross-refence of embedding performance you can check (provided as list):- "RandomFeature" will pick randomly pick input features in the same number as
latent_dimand provides a lower performance baseline. Will be repeated five times. - "PCA", "UMAP" or "TSNE" will perform dimension reduction to
latent_dimto compare autoencoder embeddings to other techniques.
- "RandomFeature" will pick randomly pick input features in the same number as
Let's check this with an advanced example:
In [4]:
Copied!
from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(n_estimators=100) # Use a custom classifier from sklearn including hyperparameters
metric_class = "accuracy" # Choose another metric to evaluate classification performance
split_type = "CV-5" # Use 5-fold cross-validation instead of the predefined train/test split
reference_methods = ["RandomFeature", "PCA"] # Compare results to additional reference methods
result = stackix.evaluate(
params=[my_target_param],
ml_model_class=rf,
metric_class=metric_class,
split_type=split_type,
reference_methods=reference_methods
)
from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(n_estimators=100) # Use a custom classifier from sklearn including hyperparameters
metric_class = "accuracy" # Choose another metric to evaluate classification performance
split_type = "CV-5" # Use 5-fold cross-validation instead of the predefined train/test split
reference_methods = ["RandomFeature", "PCA"] # Compare results to additional reference methods
result = stackix.evaluate(
params=[my_target_param],
ml_model_class=rf,
metric_class=metric_class,
split_type=split_type,
reference_methods=reference_methods
)
Perform ML task with feature df: RandomFeature RandomFeature_R1 RandomFeature_R2 RandomFeature_R3 RandomFeature_R4 RandomFeature_R5 Perform ML task with feature df: PCA PCA Perform ML task with feature df: Latent Latent Perform ML task for target parameter: condition
Note, that evaluation results will be updated with additional evaluate calls:
In [5]:
Copied!
print("Evaluation results:")
print(result.embedding_evaluation)
print("Evaluation results:")
print(result.embedding_evaluation)
Evaluation results:
score_split CLINIC_PARAM metric value ML_ALG \
0 train condition roc_auc_ovo 0.718757 LogisticRegression()
1 valid condition roc_auc_ovo 0.752485 LogisticRegression()
2 test condition roc_auc_ovo 0.673119 LogisticRegression()
0 test condition accuracy 0.700000 RandomForestClassifier()
1 test condition accuracy 0.790000 RandomForestClassifier()
.. ... ... ... ... ...
5 train condition accuracy 1.000000 RandomForestClassifier()
6 train condition accuracy 1.000000 RandomForestClassifier()
7 train condition accuracy 1.000000 RandomForestClassifier()
8 train condition accuracy 1.000000 RandomForestClassifier()
9 train condition accuracy 1.000000 RandomForestClassifier()
ML_TYPE ML_TASK ML_SUBTASK cv_run
0 classification Latent Latent NaN
1 classification Latent Latent NaN
2 classification Latent Latent NaN
0 classification RandomFeature RandomFeature_R1 CV_1
1 classification RandomFeature RandomFeature_R1 CV_2
.. ... ... ... ...
5 classification Latent Latent CV_1
6 classification Latent Latent CV_2
7 classification Latent Latent CV_3
8 classification Latent Latent CV_4
9 classification Latent Latent CV_5
[73 rows x 9 columns]
Results can be visualized as bar plots:
In [6]:
Copied!
fig_condition_eval_linear = stackix.visualizer.show_evaluation(
param=my_target_param, # The parameter to visualize
metric="accuracy", # The metric we specified
ml_alg=str(rf) # The string representation of the classifier used
)
fig_condition_eval_linear = stackix.visualizer.show_evaluation(
param=my_target_param, # The parameter to visualize
metric="accuracy", # The metric we specified
ml_alg=str(rf) # The string representation of the classifier used
)
Using evaluate with non-sklearn models¶
We can use other machine learning models which are not from sklearn, but have the same interface. An recent and powerful example is the foundation model TabPFN https://github.com/PriorLabs/TabPFN
In [7]:
Copied!
## Variable needs to set otherwise an error occurs when importing scipy
%env SCIPY_ARRAY_API=1
## Variable needs to set otherwise an error occurs when importing scipy
%env SCIPY_ARRAY_API=1
env: SCIPY_ARRAY_API=1
In [8]:
Copied!
import torch
if torch.cuda.is_available():
device = torch.cuda.current_device()
torch.cuda.get_device_properties(device).total_memory / 1024**3
if torch.cuda.is_available() and torch.cuda.get_device_properties(device).total_memory > 10*1024**4:
from tabpfn import TabPFNClassifier, TabPFNRegressor
tabpfn_class = TabPFNClassifier()
result = stackix.evaluate(
params=[my_target_param],
ml_model_class=tabpfn_class,
metric_class=metric_class,
split_type=split_type,
reference_methods=reference_methods
)
else:
print("Get a GPU with >4GB VRAM to run TabPFN models, you poor soul.")
import torch
if torch.cuda.is_available():
device = torch.cuda.current_device()
torch.cuda.get_device_properties(device).total_memory / 1024**3
if torch.cuda.is_available() and torch.cuda.get_device_properties(device).total_memory > 10*1024**4:
from tabpfn import TabPFNClassifier, TabPFNRegressor
tabpfn_class = TabPFNClassifier()
result = stackix.evaluate(
params=[my_target_param],
ml_model_class=tabpfn_class,
metric_class=metric_class,
split_type=split_type,
reference_methods=reference_methods
)
else:
print("Get a GPU with >4GB VRAM to run TabPFN models, you poor soul.")
Get a GPU with >4GB VRAM to run TabPFN models, you poor soul.
In [9]:
Copied!
if torch.cuda.is_available():
fig_condition_eval_tabpfn = stackix.visualizer.show_evaluation(
param=my_target_param, # The parameter to visualize
metric="accuracy", # The metric we specified
ml_alg=str(tabpfn_class) # The string representation of the classifier used
)
if torch.cuda.is_available():
fig_condition_eval_tabpfn = stackix.visualizer.show_evaluation(
param=my_target_param, # The parameter to visualize
metric="accuracy", # The metric we specified
ml_alg=str(tabpfn_class) # The string representation of the classifier used
)
XModalix specifics¶
Embedding evaluation¶
Since XModalix is composed of multiple VAE for each modality, evaluate behaves slightly different. In the case embedding evaluation it will perform evaluation for each VAE (modality) and test its performance for the specified downstream task.
Pure VAE comparison¶
The function xmodalix.evaluator.pure_vae_comparison() is designed to evaluate the reconstruction capability of a XModalix (reference and translated) against a normal Imagix on the image target modality. This evaluation provides insights whether the enforced latent space alignment and coupled training in the XModalix decreases or improves the reconstruction capability in comparison to image-only autoencoder.
The prerequisite for this comparison is the prior training of an imagix. An example is provided in the XModalix tutorial.