Fig. 4 and S2: paired 10x Multiome (RNA + ATAC)#
The paired RNA + chromatin accessibility analysis (Fig. 4, Supplemental Fig. S2) through UniVI’s public API. Model settings follow the archived notebook UniVI_manuscript_GR-Figure__4__Multiome_paired.ipynb (Supplemental Table S8); preprocessing follows Supplemental Table S7 using UniVI’s preprocessors, so results will be close to, not identical with, the published values.
Data: annotated 10x Multiome PBMCs, 9,631 cells (Zenodo).
import sys
if "google.colab" in sys.modules:
%pip install -q "univi[tutorials]>=1.1" "pandas==2.2.3"
import numpy as np
import pandas as pd
import scanpy as sc
import torch
import univi.datasets as uds
from univi import ModalityConfig, TrainingConfig, UniVIConfig, UniVIMultiModalVAE, UniVITrainer
from univi.evaluation import compute_modality_mixing, cross_modal_predict, encode_adata, evaluate_alignment
from univi.preprocessing import ATACPreprocessor, RNAPreprocessor, split_by_label
from univi.utils.seed import set_seed
from univi.workflows import make_loader, stack_embeddings
device = "cuda" if torch.cuda.is_available() else ("mps" if torch.backends.mps.is_available() else "cpu")
set_seed(0)
C:\Users\ashfo\micromamba\envs\univi-release-050\Lib\site-packages\tqdm\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
from .autonotebook import tqdm as notebook_tqdm
N_EPOCHS = 5000 # archived notebook: 5000 with early stopping (patience 300)
N_HVG = 2000
N_LSI = 100 # archived notebook: 100 LSI components, first component kept
Data and split#
Stratified by cell_type with at most 800 training and 100 validation cells per type (80/10 otherwise); the remaining cells form the test set (Supplemental Table S7).
data = uds.pbmc_multiome_10k()
rna, atac = data["rna"], data["atac"]
if "split" in rna.obs:
splits = {k: np.flatnonzero(rna.obs["split"].to_numpy() == k) for k in ("train", "val", "test")}
else:
splits = split_by_label(rna.obs["cell_type"], train_fraction=0.8, val_fraction=0.1,
train_cap=800, val_cap=100, seed=0)
print({k: len(v) for k, v in splits.items()})
{'train': 5777, 'val': 717, 'test': 3137}
10x-Multiome-Pbmc10k-RNA.h5ad: 0%| | 0.00/50.7M [00:00<?, ?B/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 2%|▏ | 1.05M/50.7M [00:01<01:10, 704kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 4%|▍ | 2.10M/50.7M [00:03<01:12, 667kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 6%|▌ | 3.15M/50.7M [00:04<01:09, 689kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 8%|▊ | 4.19M/50.7M [00:05<01:03, 728kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 10%|█ | 5.24M/50.7M [00:07<01:00, 754kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 12%|█▏ | 6.29M/50.7M [00:08<01:00, 738kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 14%|█▍ | 7.34M/50.7M [00:10<01:01, 704kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 17%|█▋ | 8.39M/50.7M [00:11<00:57, 733kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 19%|█▊ | 9.44M/50.7M [00:13<00:56, 727kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 21%|██ | 10.5M/50.7M [00:14<00:55, 724kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 23%|██▎ | 11.5M/50.7M [00:15<00:50, 774kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 25%|██▍ | 12.6M/50.7M [00:16<00:48, 784kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 27%|██▋ | 13.6M/50.7M [00:18<00:48, 762kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 29%|██▉ | 14.7M/50.7M [00:19<00:48, 748kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 31%|███ | 15.7M/50.7M [00:21<00:47, 738kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 33%|███▎ | 16.8M/50.7M [00:22<00:46, 730kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 35%|███▌ | 17.8M/50.7M [00:24<00:43, 751kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 37%|███▋ | 18.9M/50.7M [00:25<00:44, 714kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 39%|███▉ | 19.9M/50.7M [00:27<00:41, 739kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 41%|████▏ | 21.0M/50.7M [00:28<00:40, 732kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 43%|████▎ | 22.0M/50.7M [00:30<00:39, 727kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 46%|████▌ | 23.1M/50.7M [00:31<00:36, 749kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 48%|████▊ | 24.1M/50.7M [00:32<00:35, 739kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 50%|████▉ | 25.2M/50.7M [00:34<00:33, 757kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 52%|█████▏ | 26.2M/50.7M [00:35<00:31, 770kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 54%|█████▍ | 27.3M/50.7M [00:37<00:33, 702kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 56%|█████▌ | 28.3M/50.7M [00:38<00:30, 729kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 58%|█████▊ | 29.4M/50.7M [00:40<00:31, 678kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 60%|██████ | 30.4M/50.7M [00:41<00:29, 689kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 62%|██████▏ | 31.5M/50.7M [00:43<00:28, 671kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 64%|██████▍ | 32.5M/50.7M [00:45<00:29, 619kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 66%|██████▌ | 33.6M/50.7M [00:48<00:32, 529kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 68%|██████▊ | 34.6M/50.7M [00:49<00:28, 573kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 70%|███████ | 35.7M/50.7M [00:50<00:24, 626kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 72%|███████▏ | 36.7M/50.7M [00:52<00:21, 649kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 75%|███████▍ | 37.7M/50.7M [00:53<00:19, 666kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 77%|███████▋ | 38.8M/50.7M [00:55<00:17, 679kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 79%|███████▊ | 39.8M/50.7M [00:56<00:15, 712kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 81%|████████ | 40.9M/50.7M [00:58<00:14, 690kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 83%|████████▎ | 41.9M/50.7M [01:00<00:13, 635kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 85%|████████▍ | 43.0M/50.7M [01:01<00:12, 638kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 87%|████████▋ | 44.0M/50.7M [01:03<00:10, 659kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 89%|████████▉ | 45.1M/50.7M [01:04<00:08, 654kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 91%|█████████ | 46.1M/50.7M [01:06<00:06, 648kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 93%|█████████▎| 47.2M/50.7M [01:08<00:05, 647kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 95%|█████████▌| 48.2M/50.7M [01:09<00:03, 666kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 97%|█████████▋| 49.3M/50.7M [01:11<00:02, 680kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 99%|█████████▉| 50.3M/50.7M [01:12<00:00, 644kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 100%|██████████| 50.7M/50.7M [01:13<00:00, 650kB/s]
10x-Multiome-Pbmc10k-RNA.h5ad: 100%|██████████| 50.7M/50.7M [01:13<00:00, 690kB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 0%| | 0.00/154M [00:00<?, ?B/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 1%| | 1.05M/154M [00:02<05:08, 496kB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 1%|▏ | 2.10M/154M [00:03<03:34, 707kB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 2%|▏ | 3.15M/154M [00:03<02:06, 1.19MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 3%|▎ | 4.19M/154M [00:03<01:25, 1.75MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 3%|▎ | 5.24M/154M [00:03<01:02, 2.39MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 4%|▍ | 6.29M/154M [00:03<00:48, 3.02MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 5%|▌ | 8.39M/154M [00:03<00:30, 4.75MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 6%|▌ | 9.44M/154M [00:04<00:40, 3.60MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 7%|▋ | 10.5M/154M [00:04<00:43, 3.31MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 7%|▋ | 11.5M/154M [00:05<00:43, 3.26MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 8%|▊ | 12.6M/154M [00:05<00:43, 3.24MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 9%|▉ | 13.6M/154M [00:06<00:49, 2.82MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 10%|▉ | 14.7M/154M [00:06<00:41, 3.34MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 10%|█ | 15.7M/154M [00:06<00:48, 2.85MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 11%|█ | 16.8M/154M [00:06<00:41, 3.31MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 12%|█▏ | 17.8M/154M [00:07<00:47, 2.84MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 12%|█▏ | 18.9M/154M [00:07<00:46, 2.93MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 13%|█▎ | 19.9M/154M [00:08<00:44, 3.00MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 14%|█▎ | 21.0M/154M [00:08<00:44, 3.02MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 14%|█▍ | 22.0M/154M [00:08<00:43, 3.04MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 15%|█▍ | 23.1M/154M [00:09<00:44, 2.96MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 16%|█▌ | 24.1M/154M [00:09<00:43, 2.97MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 16%|█▋ | 25.2M/154M [00:09<00:42, 3.01MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 17%|█▋ | 26.2M/154M [00:10<00:41, 3.06MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 18%|█▊ | 27.3M/154M [00:10<00:43, 2.88MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 18%|█▊ | 28.3M/154M [00:10<00:36, 3.43MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 19%|█▉ | 29.4M/154M [00:11<00:37, 3.34MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 20%|█▉ | 30.4M/154M [00:11<00:37, 3.27MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 20%|██ | 31.5M/154M [00:11<00:38, 3.22MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 21%|██ | 32.5M/154M [00:12<00:37, 3.21MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 22%|██▏ | 33.6M/154M [00:12<00:42, 2.82MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 22%|██▏ | 34.6M/154M [00:12<00:41, 2.91MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 23%|██▎ | 35.7M/154M [00:13<00:39, 2.99MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 24%|██▍ | 36.7M/154M [00:13<00:38, 3.03MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 25%|██▍ | 37.7M/154M [00:13<00:38, 3.05MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 25%|██▌ | 38.8M/154M [00:14<00:37, 3.09MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 26%|██▌ | 39.8M/154M [00:14<00:36, 3.11MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 27%|██▋ | 40.9M/154M [00:14<00:36, 3.11MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 27%|██▋ | 41.9M/154M [00:15<00:35, 3.12MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 28%|██▊ | 43.0M/154M [00:15<00:35, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 29%|██▊ | 44.0M/154M [00:16<00:41, 2.65MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 29%|██▉ | 45.1M/154M [00:16<00:39, 2.77MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 30%|██▉ | 46.1M/154M [00:16<00:37, 2.88MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 31%|███ | 47.2M/154M [00:17<00:36, 2.96MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 31%|███▏ | 48.2M/154M [00:17<00:35, 3.01MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 32%|███▏ | 49.3M/154M [00:17<00:34, 3.04MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 33%|███▎ | 50.3M/154M [00:18<00:33, 3.09MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 33%|███▎ | 51.4M/154M [00:18<00:32, 3.11MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 34%|███▍ | 52.4M/154M [00:18<00:32, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 35%|███▍ | 53.5M/154M [00:19<00:31, 3.14MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 35%|███▌ | 54.5M/154M [00:19<00:31, 3.14MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 36%|███▌ | 55.6M/154M [00:19<00:31, 3.14MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 37%|███▋ | 56.6M/154M [00:20<00:30, 3.16MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 37%|███▋ | 57.7M/154M [00:20<00:30, 3.18MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 38%|███▊ | 58.7M/154M [00:20<00:30, 3.17MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 39%|███▉ | 59.8M/154M [00:21<00:29, 3.19MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 40%|███▉ | 60.8M/154M [00:21<00:29, 3.18MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 40%|████ | 61.9M/154M [00:21<00:29, 3.16MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 41%|████ | 62.9M/154M [00:22<00:28, 3.17MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 42%|████▏ | 64.0M/154M [00:22<00:28, 3.17MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 42%|████▏ | 65.0M/154M [00:22<00:30, 2.90MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 43%|████▎ | 66.1M/154M [00:23<00:29, 2.94MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 44%|████▎ | 67.1M/154M [00:23<00:29, 2.97MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 44%|████▍ | 68.2M/154M [00:23<00:28, 3.03MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 45%|████▍ | 69.2M/154M [00:24<00:27, 3.07MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 46%|████▌ | 70.3M/154M [00:24<00:26, 3.10MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 46%|████▋ | 71.3M/154M [00:24<00:26, 3.11MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 47%|████▋ | 72.4M/154M [00:25<00:26, 3.12MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 48%|████▊ | 73.4M/154M [00:25<00:26, 3.00MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 48%|████▊ | 74.4M/154M [00:25<00:26, 3.03MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 49%|████▉ | 75.5M/154M [00:26<00:25, 3.08MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 50%|████▉ | 76.5M/154M [00:26<00:27, 2.82MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 50%|█████ | 77.6M/154M [00:26<00:26, 2.92MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 51%|█████ | 78.6M/154M [00:27<00:25, 2.99MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 52%|█████▏ | 79.7M/154M [00:27<00:24, 3.02MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 52%|█████▏ | 80.7M/154M [00:27<00:23, 3.06MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 53%|█████▎ | 81.8M/154M [00:28<00:23, 3.08MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 54%|█████▍ | 82.8M/154M [00:28<00:23, 3.04MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 55%|█████▍ | 83.9M/154M [00:28<00:22, 3.06MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 55%|█████▌ | 84.9M/154M [00:29<00:22, 3.07MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 56%|█████▌ | 86.0M/154M [00:29<00:18, 3.61MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 57%|█████▋ | 87.0M/154M [00:29<00:19, 3.44MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 57%|█████▋ | 88.1M/154M [00:30<00:20, 3.29MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 58%|█████▊ | 89.1M/154M [00:30<00:28, 2.28MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 59%|█████▉ | 91.2M/154M [00:31<00:17, 3.65MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 61%|██████ | 93.3M/154M [00:31<00:13, 4.40MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 63%|██████▎ | 97.5M/154M [00:31<00:07, 7.77MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 65%|██████▍ | 99.6M/154M [00:32<00:10, 4.99MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 65%|██████▌ | 101M/154M [00:32<00:11, 4.57MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 66%|██████▌ | 102M/154M [00:33<00:12, 4.21MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 67%|██████▋ | 103M/154M [00:33<00:13, 3.91MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 67%|██████▋ | 104M/154M [00:33<00:15, 3.31MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 68%|██████▊ | 105M/154M [00:34<00:15, 3.26MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 69%|██████▉ | 106M/154M [00:34<00:14, 3.25MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 70%|██████▉ | 107M/154M [00:34<00:14, 3.20MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 70%|███████ | 108M/154M [00:35<00:14, 3.19MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 71%|███████ | 109M/154M [00:35<00:14, 3.18MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 72%|███████▏ | 110M/154M [00:35<00:11, 3.72MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 72%|███████▏ | 111M/154M [00:36<00:15, 2.68MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 73%|███████▎ | 112M/154M [00:36<00:13, 3.20MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 74%|███████▎ | 113M/154M [00:36<00:12, 3.17MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 74%|███████▍ | 114M/154M [00:37<00:12, 3.17MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 75%|███████▍ | 115M/154M [00:37<00:12, 3.16MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 76%|███████▌ | 116M/154M [00:38<00:13, 2.74MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 76%|███████▋ | 117M/154M [00:38<00:12, 2.83MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 77%|███████▋ | 118M/154M [00:38<00:12, 2.92MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 78%|███████▊ | 120M/154M [00:39<00:11, 3.00MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 78%|███████▊ | 121M/154M [00:39<00:10, 3.04MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 79%|███████▉ | 122M/154M [00:39<00:09, 3.58MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 80%|███████▉ | 123M/154M [00:39<00:09, 3.43MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 80%|████████ | 124M/154M [00:40<00:10, 2.91MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 81%|████████ | 125M/154M [00:40<00:08, 3.36MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 82%|████████▏ | 126M/154M [00:41<00:10, 2.74MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 82%|████████▏ | 127M/154M [00:41<00:09, 2.86MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 83%|████████▎ | 128M/154M [00:41<00:08, 2.94MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 84%|████████▍ | 129M/154M [00:42<00:08, 3.01MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 85%|████████▍ | 130M/154M [00:42<00:07, 3.05MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 85%|████████▌ | 131M/154M [00:42<00:07, 3.03MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 86%|████████▌ | 132M/154M [00:43<00:07, 3.07MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 87%|████████▋ | 133M/154M [00:43<00:06, 2.99MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 87%|████████▋ | 134M/154M [00:44<00:07, 2.68MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 88%|████████▊ | 135M/154M [00:44<00:06, 2.81MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 89%|████████▊ | 136M/154M [00:44<00:06, 2.91MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 89%|████████▉ | 137M/154M [00:45<00:05, 2.98MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 90%|████████▉ | 138M/154M [00:45<00:05, 3.02MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 91%|█████████ | 139M/154M [00:45<00:04, 3.05MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 91%|█████████▏| 141M/154M [00:46<00:04, 3.08MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 92%|█████████▏| 142M/154M [00:46<00:03, 3.09MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 93%|█████████▎| 143M/154M [00:46<00:03, 3.10MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 93%|█████████▎| 144M/154M [00:47<00:03, 3.12MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 94%|█████████▍| 145M/154M [00:47<00:02, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 95%|█████████▍| 146M/154M [00:47<00:02, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 95%|█████████▌| 147M/154M [00:48<00:02, 3.14MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 96%|█████████▌| 148M/154M [00:48<00:01, 3.15MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 97%|█████████▋| 149M/154M [00:48<00:01, 3.10MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 97%|█████████▋| 150M/154M [00:49<00:01, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 98%|█████████▊| 151M/154M [00:49<00:00, 3.12MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 99%|█████████▉| 152M/154M [00:49<00:00, 3.13MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 99%|█████████▉| 153M/154M [00:50<00:00, 3.12MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 100%|██████████| 154M/154M [00:50<00:00, 3.34MB/s]
10x-Multiome-Pbmc10k-ATAC.h5ad: 100%|██████████| 154M/154M [00:50<00:00, 3.06MB/s]
rna_prep = RNAPreprocessor(n_hvg=N_HVG, scale=True).fit(rna[splits["train"]])
atac_prep = ATACPreprocessor(n_components=N_LSI, drop_first=False, scale=True).fit(atac[splits["train"]])
parts = {k: {"rna": rna_prep.transform(rna[i]), "atac": atac_prep.transform(atac[i])} for k, i in splits.items()}
cfg = UniVIConfig(
latent_dim=30, beta=1.25, gamma=4.35, encoder_dropout=0.1, decoder_dropout=0.05,
kl_anneal_start=50, kl_anneal_end=85, align_anneal_start=75, align_anneal_end=110,
modalities=[
ModalityConfig("rna", parts["train"]["rna"].n_vars, [512, 256, 128], [128, 256, 512], likelihood="gaussian"),
ModalityConfig("atac", parts["train"]["atac"].n_vars, [128, 64], [64, 128], likelihood="gaussian"),
],
)
model = UniVIMultiModalVAE(cfg, loss_mode="v1", v1_recon="avg", normalize_v1_terms=True)
UniVITrainer(
model, make_loader(parts["train"], batch_size=128, shuffle=True, drop_last=True),
make_loader(parts["val"], batch_size=1024),
TrainingConfig(n_epochs=N_EPOCHS, batch_size=128, lr=1e-3, weight_decay=1e-4, device=device,
early_stopping=True, patience=300, best_epoch_warmup=110, log_every=100),
).fit();
[2026-09-21 14:51:51,488] [UniVITrainer] [INFO] TrainingConfig:
[2026-09-21 14:51:51,489] [UniVITrainer] [INFO] n_epochs: 5000
[2026-09-21 14:51:51,489] [UniVITrainer] [INFO] batch_size: 128
[2026-09-21 14:51:51,489] [UniVITrainer] [INFO] lr: 0.001
[2026-09-21 14:51:51,490] [UniVITrainer] [INFO] weight_decay: 0.0001
[2026-09-21 14:51:51,490] [UniVITrainer] [INFO] device: 'cuda'
[2026-09-21 14:51:51,490] [UniVITrainer] [INFO] log_every: 100
[2026-09-21 14:51:51,490] [UniVITrainer] [INFO] grad_clip: None
[2026-09-21 14:51:51,491] [UniVITrainer] [INFO] num_workers: 0
[2026-09-21 14:51:51,492] [UniVITrainer] [INFO] seed: 0
[2026-09-21 14:51:51,492] [UniVITrainer] [INFO] early_stopping: True
[2026-09-21 14:51:51,492] [UniVITrainer] [INFO] patience: 300
[2026-09-21 14:51:51,493] [UniVITrainer] [INFO] min_delta: 0.0
[2026-09-21 14:51:51,493] [UniVITrainer] [INFO] best_epoch_warmup: 110
Training UniVI: 0%| | 0/5000 [00:00<?, ?it/s]
[2026-09-21 14:51:52,618] [UniVITrainer] [INFO] [Epoch 001] Train loss=1024.6872 (beta=0.000, gamma=0.000)
[2026-09-21 14:51:52,644] [UniVITrainer] [INFO] [Epoch 001] Val loss=3815.4526 (beta=1.250, gamma=4.350)
Training UniVI: 0%| | 0/5000 [00:01<?, ?it/s, beta=0.000, gamma=0.000, train_loss=1024.6872, val_loss=3815.4526]
Training UniVI: 0%| | 1/5000 [00:01<1:35:55, 1.15s/it, beta=0.000, gamma=0.000, train_loss=1024.6872, val_loss=3815.4526]
Training UniVI: 0%| | 1/5000 [00:02<1:35:55, 1.15s/it, beta=0.000, gamma=0.000, train_loss=966.5840, val_loss=19204.7246]
Training UniVI: 0%| | 2/5000 [00:02<1:23:57, 1.01s/it, beta=0.000, gamma=0.000, train_loss=966.5840, val_loss=19204.7246]
Training UniVI: 0%| | 2/5000 [00:02<1:23:57, 1.01s/it, beta=0.000, gamma=0.000, train_loss=942.4511, val_loss=2762.1797]
Training UniVI: 0%| | 3/5000 [00:02<1:18:31, 1.06it/s, beta=0.000, gamma=0.000, train_loss=942.4511, val_loss=2762.1797]
Training UniVI: 0%| | 3/5000 [00:03<1:18:31, 1.06it/s, beta=0.000, gamma=0.000, train_loss=936.6471, val_loss=2649.6045]
Training UniVI: 0%| | 4/5000 [00:03<1:15:11, 1.11it/s, beta=0.000, gamma=0.000, train_loss=936.6471, val_loss=2649.6045]
Training UniVI: 0%| | 4/5000 [00:04<1:15:11, 1.11it/s, beta=0.000, gamma=0.000, train_loss=929.8688, val_loss=2400.1550]
Training UniVI: 0%| | 5/5000 [00:04<1:13:13, 1.14it/s, beta=0.000, gamma=0.000, train_loss=929.8688, val_loss=2400.1550]
Training UniVI: 0%| | 5/5000 [00:05<1:13:13, 1.14it/s, beta=0.000, gamma=0.000, train_loss=922.1409, val_loss=4383.9126]
Training UniVI: 0%| | 6/5000 [00:05<1:11:43, 1.16it/s, beta=0.000, gamma=0.000, train_loss=922.1409, val_loss=4383.9126]
Training UniVI: 0%| | 6/5000 [00:06<1:11:43, 1.16it/s, beta=0.000, gamma=0.000, train_loss=916.3660, val_loss=4589.2866]
Training UniVI: 0%| | 7/5000 [00:06<1:14:55, 1.11it/s, beta=0.000, gamma=0.000, train_loss=916.3660, val_loss=4589.2866]
Training UniVI: 0%| | 7/5000 [00:07<1:14:55, 1.11it/s, beta=0.000, gamma=0.000, train_loss=911.2655, val_loss=4214.0410]
Training UniVI: 0%| | 8/5000 [00:07<1:14:52, 1.11it/s, beta=0.000, gamma=0.000, train_loss=911.2655, val_loss=4214.0410]
Training UniVI: 0%| | 8/5000 [00:08<1:14:52, 1.11it/s, beta=0.000, gamma=0.000, train_loss=905.9666, val_loss=5604.7783]
Training UniVI: 0%| | 9/5000 [00:08<1:14:40, 1.11it/s, beta=0.000, gamma=0.000, train_loss=905.9666, val_loss=5604.7783]
Training UniVI: 0%| | 9/5000 [00:09<1:14:40, 1.11it/s, beta=0.000, gamma=0.000, train_loss=903.1590, val_loss=7213.5752]
Training UniVI: 0%| | 10/5000 [00:09<1:14:17, 1.12it/s, beta=0.000, gamma=0.000, train_loss=903.1590, val_loss=7213.5752]
Training UniVI: 0%| | 10/5000 [00:10<1:14:17, 1.12it/s, beta=0.000, gamma=0.000, train_loss=900.8475, val_loss=7265.1592]
Training UniVI: 0%| | 11/5000 [00:10<1:14:53, 1.11it/s, beta=0.000, gamma=0.000, train_loss=900.8475, val_loss=7265.1592]
Training UniVI: 0%| | 11/5000 [00:10<1:14:53, 1.11it/s, beta=0.000, gamma=0.000, train_loss=898.1132, val_loss=8277.2822]
Training UniVI: 0%| | 12/5000 [00:10<1:14:33, 1.12it/s, beta=0.000, gamma=0.000, train_loss=898.1132, val_loss=8277.2822]
Training UniVI: 0%| | 12/5000 [00:11<1:14:33, 1.12it/s, beta=0.000, gamma=0.000, train_loss=897.1454, val_loss=12728.1914]
Training UniVI: 0%| | 13/5000 [00:11<1:15:34, 1.10it/s, beta=0.000, gamma=0.000, train_loss=897.1454, val_loss=12728.1914]
Training UniVI: 0%| | 13/5000 [00:12<1:15:34, 1.10it/s, beta=0.000, gamma=0.000, train_loss=894.8817, val_loss=15645.7783]
Training UniVI: 0%| | 14/5000 [00:12<1:17:03, 1.08it/s, beta=0.000, gamma=0.000, train_loss=894.8817, val_loss=15645.7783]
Training UniVI: 0%| | 14/5000 [00:13<1:17:03, 1.08it/s, beta=0.000, gamma=0.000, train_loss=891.3607, val_loss=19718.6270]
Training UniVI: 0%| | 15/5000 [00:13<1:15:38, 1.10it/s, beta=0.000, gamma=0.000, train_loss=891.3607, val_loss=19718.6270]
Training UniVI: 0%| | 15/5000 [00:14<1:15:38, 1.10it/s, beta=0.000, gamma=0.000, train_loss=889.4568, val_loss=17418.4570]
Training UniVI: 0%| | 16/5000 [00:14<1:16:08, 1.09it/s, beta=0.000, gamma=0.000, train_loss=889.4568, val_loss=17418.4570]
Training UniVI: 0%| | 16/5000 [00:15<1:16:08, 1.09it/s, beta=0.000, gamma=0.000, train_loss=887.4463, val_loss=19342.5840]
Training UniVI: 0%| | 17/5000 [00:15<1:14:49, 1.11it/s, beta=0.000, gamma=0.000, train_loss=887.4463, val_loss=19342.5840]
Training UniVI: 0%| | 17/5000 [00:16<1:14:49, 1.11it/s, beta=0.000, gamma=0.000, train_loss=885.6424, val_loss=20524.3398]
Training UniVI: 0%| | 18/5000 [00:16<1:14:18, 1.12it/s, beta=0.000, gamma=0.000, train_loss=885.6424, val_loss=20524.3398]
Training UniVI: 0%| | 18/5000 [00:17<1:14:18, 1.12it/s, beta=0.000, gamma=0.000, train_loss=885.3821, val_loss=23451.7773]
Training UniVI: 0%| | 19/5000 [00:17<1:13:39, 1.13it/s, beta=0.000, gamma=0.000, train_loss=885.3821, val_loss=23451.7773]
Training UniVI: 0%| | 19/5000 [00:18<1:13:39, 1.13it/s, beta=0.000, gamma=0.000, train_loss=884.0255, val_loss=20110.7617]
Training UniVI: 0%| | 20/5000 [00:18<1:12:45, 1.14it/s, beta=0.000, gamma=0.000, train_loss=884.0255, val_loss=20110.7617]
Training UniVI: 0%| | 20/5000 [00:18<1:12:45, 1.14it/s, beta=0.000, gamma=0.000, train_loss=881.1856, val_loss=28020.2559]
Training UniVI: 0%| | 21/5000 [00:18<1:13:29, 1.13it/s, beta=0.000, gamma=0.000, train_loss=881.1856, val_loss=28020.2559]
Training UniVI: 0%| | 21/5000 [00:19<1:13:29, 1.13it/s, beta=0.000, gamma=0.000, train_loss=879.3244, val_loss=27589.9492]
Training UniVI: 0%| | 22/5000 [00:19<1:13:58, 1.12it/s, beta=0.000, gamma=0.000, train_loss=879.3244, val_loss=27589.9492]
Training UniVI: 0%| | 22/5000 [00:20<1:13:58, 1.12it/s, beta=0.000, gamma=0.000, train_loss=880.9321, val_loss=31437.3340]
Training UniVI: 0%| | 23/5000 [00:20<1:14:23, 1.12it/s, beta=0.000, gamma=0.000, train_loss=880.9321, val_loss=31437.3340]
Training UniVI: 0%| | 23/5000 [00:21<1:14:23, 1.12it/s, beta=0.000, gamma=0.000, train_loss=879.9125, val_loss=33391.7188]
Training UniVI: 0%| | 24/5000 [00:21<1:14:31, 1.11it/s, beta=0.000, gamma=0.000, train_loss=879.9125, val_loss=33391.7188]
Training UniVI: 0%| | 24/5000 [00:22<1:14:31, 1.11it/s, beta=0.000, gamma=0.000, train_loss=879.0638, val_loss=44209.4219]
Training UniVI: 0%| | 25/5000 [00:22<1:15:24, 1.10it/s, beta=0.000, gamma=0.000, train_loss=879.0638, val_loss=44209.4219]
Training UniVI: 0%| | 25/5000 [00:23<1:15:24, 1.10it/s, beta=0.000, gamma=0.000, train_loss=878.4670, val_loss=40984.1016]
Training UniVI: 1%| | 26/5000 [00:23<1:13:41, 1.13it/s, beta=0.000, gamma=0.000, train_loss=878.4670, val_loss=40984.1016]
Training UniVI: 1%| | 26/5000 [00:24<1:13:41, 1.13it/s, beta=0.000, gamma=0.000, train_loss=876.4828, val_loss=45284.3164]
Training UniVI: 1%| | 27/5000 [00:24<1:14:44, 1.11it/s, beta=0.000, gamma=0.000, train_loss=876.4828, val_loss=45284.3164]
Training UniVI: 1%| | 27/5000 [00:25<1:14:44, 1.11it/s, beta=0.000, gamma=0.000, train_loss=876.9056, val_loss=59022.7070]
Training UniVI: 1%| | 28/5000 [00:25<1:15:32, 1.10it/s, beta=0.000, gamma=0.000, train_loss=876.9056, val_loss=59022.7070]
Training UniVI: 1%| | 28/5000 [00:26<1:15:32, 1.10it/s, beta=0.000, gamma=0.000, train_loss=875.7910, val_loss=64534.4062]
Training UniVI: 1%| | 29/5000 [00:26<1:13:58, 1.12it/s, beta=0.000, gamma=0.000, train_loss=875.7910, val_loss=64534.4062]
Training UniVI: 1%| | 29/5000 [00:27<1:13:58, 1.12it/s, beta=0.000, gamma=0.000, train_loss=875.8087, val_loss=64437.1680]
Training UniVI: 1%| | 30/5000 [00:27<1:13:43, 1.12it/s, beta=0.000, gamma=0.000, train_loss=875.8087, val_loss=64437.1680]
Training UniVI: 1%| | 30/5000 [00:27<1:13:43, 1.12it/s, beta=0.000, gamma=0.000, train_loss=874.5301, val_loss=70733.7656]
Training UniVI: 1%| | 31/5000 [00:27<1:12:37, 1.14it/s, beta=0.000, gamma=0.000, train_loss=874.5301, val_loss=70733.7656]
Training UniVI: 1%| | 31/5000 [00:28<1:12:37, 1.14it/s, beta=0.000, gamma=0.000, train_loss=874.5295, val_loss=78898.6406]
Training UniVI: 1%| | 32/5000 [00:28<1:13:11, 1.13it/s, beta=0.000, gamma=0.000, train_loss=874.5295, val_loss=78898.6406]
Training UniVI: 1%| | 32/5000 [00:29<1:13:11, 1.13it/s, beta=0.000, gamma=0.000, train_loss=873.2712, val_loss=91814.2344]
Training UniVI: 1%| | 33/5000 [00:29<1:14:00, 1.12it/s, beta=0.000, gamma=0.000, train_loss=873.2712, val_loss=91814.2344]
Training UniVI: 1%| | 33/5000 [00:30<1:14:00, 1.12it/s, beta=0.000, gamma=0.000, train_loss=873.2822, val_loss=102181.9688]
Training UniVI: 1%| | 34/5000 [00:30<1:15:39, 1.09it/s, beta=0.000, gamma=0.000, train_loss=873.2822, val_loss=102181.9688]
Training UniVI: 1%| | 34/5000 [00:31<1:15:39, 1.09it/s, beta=0.000, gamma=0.000, train_loss=872.9351, val_loss=116610.1641]
Training UniVI: 1%| | 35/5000 [00:31<1:16:35, 1.08it/s, beta=0.000, gamma=0.000, train_loss=872.9351, val_loss=116610.1641]
Training UniVI: 1%| | 35/5000 [00:32<1:16:35, 1.08it/s, beta=0.000, gamma=0.000, train_loss=871.5273, val_loss=129629.9219]
Training UniVI: 1%| | 36/5000 [00:32<1:17:18, 1.07it/s, beta=0.000, gamma=0.000, train_loss=871.5273, val_loss=129629.9219]
Training UniVI: 1%| | 36/5000 [00:33<1:17:18, 1.07it/s, beta=0.000, gamma=0.000, train_loss=870.7913, val_loss=124192.0078]
Training UniVI: 1%| | 37/5000 [00:33<1:18:21, 1.06it/s, beta=0.000, gamma=0.000, train_loss=870.7913, val_loss=124192.0078]
Training UniVI: 1%| | 37/5000 [00:34<1:18:21, 1.06it/s, beta=0.000, gamma=0.000, train_loss=870.8937, val_loss=144797.0312]
Training UniVI: 1%| | 38/5000 [00:34<1:16:28, 1.08it/s, beta=0.000, gamma=0.000, train_loss=870.8937, val_loss=144797.0312]
Training UniVI: 1%| | 38/5000 [00:35<1:16:28, 1.08it/s, beta=0.000, gamma=0.000, train_loss=870.2668, val_loss=146431.1562]
Training UniVI: 1%| | 39/5000 [00:35<1:15:14, 1.10it/s, beta=0.000, gamma=0.000, train_loss=870.2668, val_loss=146431.1562]
Training UniVI: 1%| | 39/5000 [00:36<1:15:14, 1.10it/s, beta=0.000, gamma=0.000, train_loss=869.7292, val_loss=153240.6719]
Training UniVI: 1%| | 40/5000 [00:36<1:14:52, 1.10it/s, beta=0.000, gamma=0.000, train_loss=869.7292, val_loss=153240.6719]
Training UniVI: 1%| | 40/5000 [00:37<1:14:52, 1.10it/s, beta=0.000, gamma=0.000, train_loss=868.5496, val_loss=165260.6406]
Training UniVI: 1%| | 41/5000 [00:37<1:16:53, 1.07it/s, beta=0.000, gamma=0.000, train_loss=868.5496, val_loss=165260.6406]
Training UniVI: 1%| | 41/5000 [00:38<1:16:53, 1.07it/s, beta=0.000, gamma=0.000, train_loss=867.2160, val_loss=173342.3594]
Training UniVI: 1%| | 42/5000 [00:38<1:14:39, 1.11it/s, beta=0.000, gamma=0.000, train_loss=867.2160, val_loss=173342.3594]
Training UniVI: 1%| | 42/5000 [00:38<1:14:39, 1.11it/s, beta=0.000, gamma=0.000, train_loss=868.1997, val_loss=191284.7812]
Training UniVI: 1%| | 43/5000 [00:38<1:12:27, 1.14it/s, beta=0.000, gamma=0.000, train_loss=868.1997, val_loss=191284.7812]
Training UniVI: 1%| | 43/5000 [00:39<1:12:27, 1.14it/s, beta=0.000, gamma=0.000, train_loss=867.7604, val_loss=198119.2344]
Training UniVI: 1%| | 44/5000 [00:39<1:13:08, 1.13it/s, beta=0.000, gamma=0.000, train_loss=867.7604, val_loss=198119.2344]
Training UniVI: 1%| | 44/5000 [00:40<1:13:08, 1.13it/s, beta=0.000, gamma=0.000, train_loss=865.9936, val_loss=224431.4844]
Training UniVI: 1%| | 45/5000 [00:40<1:12:14, 1.14it/s, beta=0.000, gamma=0.000, train_loss=865.9936, val_loss=224431.4844]
Training UniVI: 1%| | 45/5000 [00:41<1:12:14, 1.14it/s, beta=0.000, gamma=0.000, train_loss=865.5041, val_loss=224090.6719]
Training UniVI: 1%| | 46/5000 [00:41<1:12:43, 1.14it/s, beta=0.000, gamma=0.000, train_loss=865.5041, val_loss=224090.6719]
Training UniVI: 1%| | 46/5000 [00:42<1:12:43, 1.14it/s, beta=0.000, gamma=0.000, train_loss=864.8479, val_loss=243196.4375]
Training UniVI: 1%| | 47/5000 [00:42<1:14:08, 1.11it/s, beta=0.000, gamma=0.000, train_loss=864.8479, val_loss=243196.4375]
Training UniVI: 1%| | 47/5000 [00:43<1:14:08, 1.11it/s, beta=0.000, gamma=0.000, train_loss=864.8118, val_loss=269186.2188]
Training UniVI: 1%| | 48/5000 [00:43<1:17:18, 1.07it/s, beta=0.000, gamma=0.000, train_loss=864.8118, val_loss=269186.2188]
Training UniVI: 1%| | 48/5000 [00:44<1:17:18, 1.07it/s, beta=0.000, gamma=0.000, train_loss=863.9688, val_loss=265283.0000]
Training UniVI: 1%| | 49/5000 [00:44<1:14:55, 1.10it/s, beta=0.000, gamma=0.000, train_loss=863.9688, val_loss=265283.0000]
Training UniVI: 1%| | 49/5000 [00:45<1:14:55, 1.10it/s, beta=0.000, gamma=0.000, train_loss=863.2635, val_loss=291739.8438]
Training UniVI: 1%| | 50/5000 [00:45<1:14:08, 1.11it/s, beta=0.000, gamma=0.000, train_loss=863.2635, val_loss=291739.8438]
Training UniVI: 1%| | 50/5000 [00:46<1:14:08, 1.11it/s, beta=0.036, gamma=0.000, train_loss=867.2224, val_loss=46794.2852]
Training UniVI: 1%| | 51/5000 [00:46<1:15:29, 1.09it/s, beta=0.036, gamma=0.000, train_loss=867.2224, val_loss=46794.2852]
Training UniVI: 1%| | 51/5000 [00:47<1:15:29, 1.09it/s, beta=0.071, gamma=0.000, train_loss=869.2299, val_loss=4176.6494]
Training UniVI: 1%| | 52/5000 [00:47<1:14:49, 1.10it/s, beta=0.071, gamma=0.000, train_loss=869.2299, val_loss=4176.6494]
Training UniVI: 1%| | 52/5000 [00:48<1:14:49, 1.10it/s, beta=0.107, gamma=0.000, train_loss=869.4642, val_loss=1606.1510]
Training UniVI: 1%| | 53/5000 [00:48<1:16:18, 1.08it/s, beta=0.107, gamma=0.000, train_loss=869.4642, val_loss=1606.1510]
Training UniVI: 1%| | 53/5000 [00:48<1:16:18, 1.08it/s, beta=0.143, gamma=0.000, train_loss=869.6278, val_loss=1319.6672]
Training UniVI: 1%| | 54/5000 [00:48<1:14:50, 1.10it/s, beta=0.143, gamma=0.000, train_loss=869.6278, val_loss=1319.6672]
Training UniVI: 1%| | 54/5000 [00:49<1:14:50, 1.10it/s, beta=0.179, gamma=0.000, train_loss=871.0731, val_loss=1201.0836]
Training UniVI: 1%| | 55/5000 [00:49<1:14:18, 1.11it/s, beta=0.179, gamma=0.000, train_loss=871.0731, val_loss=1201.0836]
Training UniVI: 1%| | 55/5000 [00:50<1:14:18, 1.11it/s, beta=0.214, gamma=0.000, train_loss=871.4625, val_loss=1170.5276]
Training UniVI: 1%| | 56/5000 [00:50<1:12:38, 1.13it/s, beta=0.214, gamma=0.000, train_loss=871.4625, val_loss=1170.5276]
Training UniVI: 1%| | 56/5000 [00:51<1:12:38, 1.13it/s, beta=0.250, gamma=0.000, train_loss=872.6165, val_loss=1105.9008]
Training UniVI: 1%| | 57/5000 [00:51<1:11:42, 1.15it/s, beta=0.250, gamma=0.000, train_loss=872.6165, val_loss=1105.9008]
Training UniVI: 1%| | 57/5000 [00:52<1:11:42, 1.15it/s, beta=0.286, gamma=0.000, train_loss=873.2953, val_loss=1087.4467]
Training UniVI: 1%| | 58/5000 [00:52<1:12:42, 1.13it/s, beta=0.286, gamma=0.000, train_loss=873.2953, val_loss=1087.4467]
Training UniVI: 1%| | 58/5000 [00:53<1:12:42, 1.13it/s, beta=0.321, gamma=0.000, train_loss=873.1899, val_loss=1057.7310]
Training UniVI: 1%| | 59/5000 [00:53<1:12:00, 1.14it/s, beta=0.321, gamma=0.000, train_loss=873.1899, val_loss=1057.7310]
Training UniVI: 1%| | 59/5000 [00:54<1:12:00, 1.14it/s, beta=0.357, gamma=0.000, train_loss=873.9506, val_loss=1039.8494]
Training UniVI: 1%| | 60/5000 [00:54<1:11:38, 1.15it/s, beta=0.357, gamma=0.000, train_loss=873.9506, val_loss=1039.8494]
Training UniVI: 1%| | 60/5000 [00:54<1:11:38, 1.15it/s, beta=0.393, gamma=0.000, train_loss=873.1831, val_loss=1031.1863]
Training UniVI: 1%| | 61/5000 [00:54<1:11:36, 1.15it/s, beta=0.393, gamma=0.000, train_loss=873.1831, val_loss=1031.1863]
Training UniVI: 1%| | 61/5000 [00:55<1:11:36, 1.15it/s, beta=0.429, gamma=0.000, train_loss=873.4069, val_loss=1017.0833]
Training UniVI: 1%| | 62/5000 [00:55<1:11:06, 1.16it/s, beta=0.429, gamma=0.000, train_loss=873.4069, val_loss=1017.0833]
Training UniVI: 1%| | 62/5000 [00:56<1:11:06, 1.16it/s, beta=0.464, gamma=0.000, train_loss=873.6329, val_loss=1013.8430]
Training UniVI: 1%|▏ | 63/5000 [00:56<1:11:32, 1.15it/s, beta=0.464, gamma=0.000, train_loss=873.6329, val_loss=1013.8430]
Training UniVI: 1%|▏ | 63/5000 [00:57<1:11:32, 1.15it/s, beta=0.500, gamma=0.000, train_loss=874.1239, val_loss=998.0175]
Training UniVI: 1%|▏ | 64/5000 [00:57<1:12:08, 1.14it/s, beta=0.500, gamma=0.000, train_loss=874.1239, val_loss=998.0175]
Training UniVI: 1%|▏ | 64/5000 [00:58<1:12:08, 1.14it/s, beta=0.536, gamma=0.000, train_loss=874.3535, val_loss=992.2283]
Training UniVI: 1%|▏ | 65/5000 [00:58<1:11:22, 1.15it/s, beta=0.536, gamma=0.000, train_loss=874.3535, val_loss=992.2283]
Training UniVI: 1%|▏ | 65/5000 [00:59<1:11:22, 1.15it/s, beta=0.571, gamma=0.000, train_loss=873.4619, val_loss=984.3943]
Training UniVI: 1%|▏ | 66/5000 [00:59<1:10:45, 1.16it/s, beta=0.571, gamma=0.000, train_loss=873.4619, val_loss=984.3943]
Training UniVI: 1%|▏ | 66/5000 [01:00<1:10:45, 1.16it/s, beta=0.607, gamma=0.000, train_loss=874.5224, val_loss=974.3118]
Training UniVI: 1%|▏ | 67/5000 [01:00<1:11:09, 1.16it/s, beta=0.607, gamma=0.000, train_loss=874.5224, val_loss=974.3118]
Training UniVI: 1%|▏ | 67/5000 [01:00<1:11:09, 1.16it/s, beta=0.643, gamma=0.000, train_loss=873.5932, val_loss=972.1347]
Training UniVI: 1%|▏ | 68/5000 [01:00<1:11:08, 1.16it/s, beta=0.643, gamma=0.000, train_loss=873.5932, val_loss=972.1347]
Training UniVI: 1%|▏ | 68/5000 [01:01<1:11:08, 1.16it/s, beta=0.679, gamma=0.000, train_loss=874.6688, val_loss=966.4886]
Training UniVI: 1%|▏ | 69/5000 [01:01<1:11:21, 1.15it/s, beta=0.679, gamma=0.000, train_loss=874.6688, val_loss=966.4886]
Training UniVI: 1%|▏ | 69/5000 [01:02<1:11:21, 1.15it/s, beta=0.714, gamma=0.000, train_loss=875.3252, val_loss=967.3071]
Training UniVI: 1%|▏ | 70/5000 [01:02<1:11:11, 1.15it/s, beta=0.714, gamma=0.000, train_loss=875.3252, val_loss=967.3071]
Training UniVI: 1%|▏ | 70/5000 [01:03<1:11:11, 1.15it/s, beta=0.750, gamma=0.000, train_loss=874.7202, val_loss=960.2215]
Training UniVI: 1%|▏ | 71/5000 [01:03<1:11:12, 1.15it/s, beta=0.750, gamma=0.000, train_loss=874.7202, val_loss=960.2215]
Training UniVI: 1%|▏ | 71/5000 [01:04<1:11:12, 1.15it/s, beta=0.786, gamma=0.000, train_loss=876.4765, val_loss=963.2361]
Training UniVI: 1%|▏ | 72/5000 [01:04<1:12:29, 1.13it/s, beta=0.786, gamma=0.000, train_loss=876.4765, val_loss=963.2361]
Training UniVI: 1%|▏ | 72/5000 [01:05<1:12:29, 1.13it/s, beta=0.821, gamma=0.000, train_loss=875.3104, val_loss=962.2193]
Training UniVI: 1%|▏ | 73/5000 [01:05<1:13:42, 1.11it/s, beta=0.821, gamma=0.000, train_loss=875.3104, val_loss=962.2193]
Training UniVI: 1%|▏ | 73/5000 [01:06<1:13:42, 1.11it/s, beta=0.857, gamma=0.000, train_loss=877.5762, val_loss=961.0914]
Training UniVI: 1%|▏ | 74/5000 [01:06<1:15:18, 1.09it/s, beta=0.857, gamma=0.000, train_loss=877.5762, val_loss=961.0914]
Training UniVI: 1%|▏ | 74/5000 [01:07<1:15:18, 1.09it/s, beta=0.893, gamma=0.000, train_loss=876.7579, val_loss=955.7567]
Training UniVI: 2%|▏ | 75/5000 [01:07<1:16:29, 1.07it/s, beta=0.893, gamma=0.000, train_loss=876.7579, val_loss=955.7567]
Training UniVI: 2%|▏ | 75/5000 [01:08<1:16:29, 1.07it/s, beta=0.929, gamma=0.124, train_loss=878.4692, val_loss=930.0569]
Training UniVI: 2%|▏ | 76/5000 [01:08<1:14:26, 1.10it/s, beta=0.929, gamma=0.124, train_loss=878.4692, val_loss=930.0569]
Training UniVI: 2%|▏ | 76/5000 [01:09<1:14:26, 1.10it/s, beta=0.964, gamma=0.249, train_loss=880.1319, val_loss=918.1042]
Training UniVI: 2%|▏ | 77/5000 [01:09<1:13:56, 1.11it/s, beta=0.964, gamma=0.249, train_loss=880.1319, val_loss=918.1042]
Training UniVI: 2%|▏ | 77/5000 [01:09<1:13:56, 1.11it/s, beta=1.000, gamma=0.373, train_loss=881.2213, val_loss=914.1960]
Training UniVI: 2%|▏ | 78/5000 [01:09<1:12:53, 1.13it/s, beta=1.000, gamma=0.373, train_loss=881.2213, val_loss=914.1960]
Training UniVI: 2%|▏ | 78/5000 [01:10<1:12:53, 1.13it/s, beta=1.036, gamma=0.497, train_loss=882.5791, val_loss=911.9566]
Training UniVI: 2%|▏ | 79/5000 [01:10<1:13:29, 1.12it/s, beta=1.036, gamma=0.497, train_loss=882.5791, val_loss=911.9566]
Training UniVI: 2%|▏ | 79/5000 [01:11<1:13:29, 1.12it/s, beta=1.071, gamma=0.621, train_loss=881.9215, val_loss=904.8579]
Training UniVI: 2%|▏ | 80/5000 [01:11<1:12:31, 1.13it/s, beta=1.071, gamma=0.621, train_loss=881.9215, val_loss=904.8579]
Training UniVI: 2%|▏ | 80/5000 [01:12<1:12:31, 1.13it/s, beta=1.107, gamma=0.746, train_loss=884.1177, val_loss=905.5533]
Training UniVI: 2%|▏ | 81/5000 [01:12<1:14:31, 1.10it/s, beta=1.107, gamma=0.746, train_loss=884.1177, val_loss=905.5533]
Training UniVI: 2%|▏ | 81/5000 [01:13<1:14:31, 1.10it/s, beta=1.143, gamma=0.870, train_loss=884.0129, val_loss=903.0214]
Training UniVI: 2%|▏ | 82/5000 [01:13<1:17:02, 1.06it/s, beta=1.143, gamma=0.870, train_loss=884.0129, val_loss=903.0214]
Training UniVI: 2%|▏ | 82/5000 [01:14<1:17:02, 1.06it/s, beta=1.179, gamma=0.994, train_loss=884.7362, val_loss=900.3695]
Training UniVI: 2%|▏ | 83/5000 [01:14<1:17:06, 1.06it/s, beta=1.179, gamma=0.994, train_loss=884.7362, val_loss=900.3695]
Training UniVI: 2%|▏ | 83/5000 [01:15<1:17:06, 1.06it/s, beta=1.214, gamma=1.119, train_loss=885.9777, val_loss=900.4410]
Training UniVI: 2%|▏ | 84/5000 [01:15<1:17:11, 1.06it/s, beta=1.214, gamma=1.119, train_loss=885.9777, val_loss=900.4410]
Training UniVI: 2%|▏ | 84/5000 [01:16<1:17:11, 1.06it/s, beta=1.250, gamma=1.243, train_loss=887.1017, val_loss=898.4990]
Training UniVI: 2%|▏ | 85/5000 [01:16<1:16:01, 1.08it/s, beta=1.250, gamma=1.243, train_loss=887.1017, val_loss=898.4990]
Training UniVI: 2%|▏ | 85/5000 [01:17<1:16:01, 1.08it/s, beta=1.250, gamma=1.367, train_loss=887.3523, val_loss=896.5593]
Training UniVI: 2%|▏ | 86/5000 [01:17<1:15:49, 1.08it/s, beta=1.250, gamma=1.367, train_loss=887.3523, val_loss=896.5593]
Training UniVI: 2%|▏ | 86/5000 [01:18<1:15:49, 1.08it/s, beta=1.250, gamma=1.491, train_loss=885.5111, val_loss=895.1370]
Training UniVI: 2%|▏ | 87/5000 [01:18<1:15:58, 1.08it/s, beta=1.250, gamma=1.491, train_loss=885.5111, val_loss=895.1370]
Training UniVI: 2%|▏ | 87/5000 [01:19<1:15:58, 1.08it/s, beta=1.250, gamma=1.616, train_loss=886.1782, val_loss=895.4114]
Training UniVI: 2%|▏ | 88/5000 [01:19<1:15:11, 1.09it/s, beta=1.250, gamma=1.616, train_loss=886.1782, val_loss=895.4114]
Training UniVI: 2%|▏ | 88/5000 [01:20<1:15:11, 1.09it/s, beta=1.250, gamma=1.740, train_loss=885.7402, val_loss=895.5466]
Training UniVI: 2%|▏ | 89/5000 [01:20<1:15:54, 1.08it/s, beta=1.250, gamma=1.740, train_loss=885.7402, val_loss=895.5466]
Training UniVI: 2%|▏ | 89/5000 [01:21<1:15:54, 1.08it/s, beta=1.250, gamma=1.864, train_loss=886.9033, val_loss=894.1677]
Training UniVI: 2%|▏ | 90/5000 [01:21<1:16:36, 1.07it/s, beta=1.250, gamma=1.864, train_loss=886.9033, val_loss=894.1677]
Training UniVI: 2%|▏ | 90/5000 [01:22<1:16:36, 1.07it/s, beta=1.250, gamma=1.989, train_loss=886.2845, val_loss=892.6376]
Training UniVI: 2%|▏ | 91/5000 [01:22<1:17:05, 1.06it/s, beta=1.250, gamma=1.989, train_loss=886.2845, val_loss=892.6376]
Training UniVI: 2%|▏ | 91/5000 [01:23<1:17:05, 1.06it/s, beta=1.250, gamma=2.113, train_loss=886.5886, val_loss=892.9245]
Training UniVI: 2%|▏ | 92/5000 [01:23<1:16:45, 1.07it/s, beta=1.250, gamma=2.113, train_loss=886.5886, val_loss=892.9245]
Training UniVI: 2%|▏ | 92/5000 [01:23<1:16:45, 1.07it/s, beta=1.250, gamma=2.237, train_loss=886.6498, val_loss=891.0865]
Training UniVI: 2%|▏ | 93/5000 [01:23<1:15:00, 1.09it/s, beta=1.250, gamma=2.237, train_loss=886.6498, val_loss=891.0865]
Training UniVI: 2%|▏ | 93/5000 [01:24<1:15:00, 1.09it/s, beta=1.250, gamma=2.361, train_loss=886.9694, val_loss=892.1054]
Training UniVI: 2%|▏ | 94/5000 [01:24<1:15:06, 1.09it/s, beta=1.250, gamma=2.361, train_loss=886.9694, val_loss=892.1054]
Training UniVI: 2%|▏ | 94/5000 [01:25<1:15:06, 1.09it/s, beta=1.250, gamma=2.486, train_loss=886.5237, val_loss=891.4510]
Training UniVI: 2%|▏ | 95/5000 [01:25<1:13:30, 1.11it/s, beta=1.250, gamma=2.486, train_loss=886.5237, val_loss=891.4510]
Training UniVI: 2%|▏ | 95/5000 [01:26<1:13:30, 1.11it/s, beta=1.250, gamma=2.610, train_loss=886.7031, val_loss=890.0936]
Training UniVI: 2%|▏ | 96/5000 [01:26<1:12:50, 1.12it/s, beta=1.250, gamma=2.610, train_loss=886.7031, val_loss=890.0936]
Training UniVI: 2%|▏ | 96/5000 [01:27<1:12:50, 1.12it/s, beta=1.250, gamma=2.734, train_loss=885.0785, val_loss=889.5030]
Training UniVI: 2%|▏ | 97/5000 [01:27<1:12:43, 1.12it/s, beta=1.250, gamma=2.734, train_loss=885.0785, val_loss=889.5030]
Training UniVI: 2%|▏ | 97/5000 [01:28<1:12:43, 1.12it/s, beta=1.250, gamma=2.859, train_loss=885.7832, val_loss=890.6769]
Training UniVI: 2%|▏ | 98/5000 [01:28<1:14:04, 1.10it/s, beta=1.250, gamma=2.859, train_loss=885.7832, val_loss=890.6769]
Training UniVI: 2%|▏ | 98/5000 [01:29<1:14:04, 1.10it/s, beta=1.250, gamma=2.983, train_loss=885.6213, val_loss=889.9172]
Training UniVI: 2%|▏ | 99/5000 [01:29<1:18:10, 1.04it/s, beta=1.250, gamma=2.983, train_loss=885.6213, val_loss=889.9172]
[2026-09-21 14:53:21,852] [UniVITrainer] [INFO] [Epoch 100] Train loss=885.3058 (beta=1.250, gamma=3.107)
[2026-09-21 14:53:21,878] [UniVITrainer] [INFO] [Epoch 100] Val loss=889.2094 (beta=1.250, gamma=4.350)
Training UniVI: 2%|▏ | 99/5000 [01:30<1:18:10, 1.04it/s, beta=1.250, gamma=3.107, train_loss=885.3058, val_loss=889.2094]
Training UniVI: 2%|▏ | 100/5000 [01:30<1:17:00, 1.06it/s, beta=1.250, gamma=3.107, train_loss=885.3058, val_loss=889.2094]
Training UniVI: 2%|▏ | 100/5000 [01:31<1:17:00, 1.06it/s, beta=1.250, gamma=3.231, train_loss=887.2053, val_loss=890.5324]
Training UniVI: 2%|▏ | 101/5000 [01:31<1:16:14, 1.07it/s, beta=1.250, gamma=3.231, train_loss=887.2053, val_loss=890.5324]
Training UniVI: 2%|▏ | 101/5000 [01:32<1:16:14, 1.07it/s, beta=1.250, gamma=3.356, train_loss=886.8483, val_loss=890.0574]
Training UniVI: 2%|▏ | 102/5000 [01:32<1:15:33, 1.08it/s, beta=1.250, gamma=3.356, train_loss=886.8483, val_loss=890.0574]
Training UniVI: 2%|▏ | 102/5000 [01:33<1:15:33, 1.08it/s, beta=1.250, gamma=3.480, train_loss=885.9036, val_loss=889.7137]
Training UniVI: 2%|▏ | 103/5000 [01:33<1:15:35, 1.08it/s, beta=1.250, gamma=3.480, train_loss=885.9036, val_loss=889.7137]
Training UniVI: 2%|▏ | 103/5000 [01:34<1:15:35, 1.08it/s, beta=1.250, gamma=3.604, train_loss=887.0419, val_loss=887.8441]
Training UniVI: 2%|▏ | 104/5000 [01:34<1:14:12, 1.10it/s, beta=1.250, gamma=3.604, train_loss=887.0419, val_loss=887.8441]
Training UniVI: 2%|▏ | 104/5000 [01:34<1:14:12, 1.10it/s, beta=1.250, gamma=3.729, train_loss=885.8093, val_loss=887.9031]
Training UniVI: 2%|▏ | 105/5000 [01:34<1:13:07, 1.12it/s, beta=1.250, gamma=3.729, train_loss=885.8093, val_loss=887.9031]
Training UniVI: 2%|▏ | 105/5000 [01:35<1:13:07, 1.12it/s, beta=1.250, gamma=3.853, train_loss=886.2085, val_loss=888.8797]
Training UniVI: 2%|▏ | 106/5000 [01:35<1:13:04, 1.12it/s, beta=1.250, gamma=3.853, train_loss=886.2085, val_loss=888.8797]
Training UniVI: 2%|▏ | 106/5000 [01:36<1:13:04, 1.12it/s, beta=1.250, gamma=3.977, train_loss=885.4462, val_loss=888.4316]
Training UniVI: 2%|▏ | 107/5000 [01:36<1:15:24, 1.08it/s, beta=1.250, gamma=3.977, train_loss=885.4462, val_loss=888.4316]
Training UniVI: 2%|▏ | 107/5000 [01:37<1:15:24, 1.08it/s, beta=1.250, gamma=4.101, train_loss=886.8388, val_loss=889.7350]
Training UniVI: 2%|▏ | 108/5000 [01:37<1:15:06, 1.09it/s, beta=1.250, gamma=4.101, train_loss=886.8388, val_loss=889.7350]
Training UniVI: 2%|▏ | 108/5000 [01:38<1:15:06, 1.09it/s, beta=1.250, gamma=4.226, train_loss=888.0282, val_loss=889.1305]
[2026-09-21 14:53:30,043] [UniVITrainer] [INFO] [Epoch 109] Best tracking warmup ends next epoch (best_epoch_warmup=110).
Training UniVI: 2%|▏ | 109/5000 [01:38<1:14:08, 1.10it/s, beta=1.250, gamma=4.226, train_loss=888.0282, val_loss=889.1305]
Training UniVI: 2%|▏ | 109/5000 [01:39<1:14:08, 1.10it/s, beta=1.250, gamma=4.350, train_loss=885.9135, val_loss=887.9442]
[2026-09-21 14:53:30,930] [UniVITrainer] [INFO] [Epoch 110] New best val loss: 887.9442
Training UniVI: 2%|▏ | 110/5000 [01:39<1:13:35, 1.11it/s, beta=1.250, gamma=4.350, train_loss=885.9135, val_loss=887.9442]
Training UniVI: 2%|▏ | 110/5000 [01:40<1:13:35, 1.11it/s, beta=1.250, gamma=4.350, train_loss=885.0957, val_loss=888.5680]
Training UniVI: 2%|▏ | 111/5000 [01:40<1:13:17, 1.11it/s, beta=1.250, gamma=4.350, train_loss=885.0957, val_loss=888.5680]
Training UniVI: 2%|▏ | 111/5000 [01:41<1:13:17, 1.11it/s, beta=1.250, gamma=4.350, train_loss=885.6229, val_loss=887.1743]
[2026-09-21 14:53:32,652] [UniVITrainer] [INFO] [Epoch 112] New best val loss: 887.1743
Training UniVI: 2%|▏ | 112/5000 [01:41<1:11:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=885.6229, val_loss=887.1743]
Training UniVI: 2%|▏ | 112/5000 [01:42<1:11:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=884.6112, val_loss=888.5977]
Training UniVI: 2%|▏ | 113/5000 [01:42<1:12:12, 1.13it/s, beta=1.250, gamma=4.350, train_loss=884.6112, val_loss=888.5977]
Training UniVI: 2%|▏ | 113/5000 [01:42<1:12:12, 1.13it/s, beta=1.250, gamma=4.350, train_loss=884.3573, val_loss=888.2190]
Training UniVI: 2%|▏ | 114/5000 [01:42<1:11:09, 1.14it/s, beta=1.250, gamma=4.350, train_loss=884.3573, val_loss=888.2190]
Training UniVI: 2%|▏ | 114/5000 [01:43<1:11:09, 1.14it/s, beta=1.250, gamma=4.350, train_loss=883.5718, val_loss=890.5065]
Training UniVI: 2%|▏ | 115/5000 [01:43<1:12:32, 1.12it/s, beta=1.250, gamma=4.350, train_loss=883.5718, val_loss=890.5065]
Training UniVI: 2%|▏ | 115/5000 [01:44<1:12:32, 1.12it/s, beta=1.250, gamma=4.350, train_loss=883.8950, val_loss=889.2964]
Training UniVI: 2%|▏ | 116/5000 [01:44<1:12:50, 1.12it/s, beta=1.250, gamma=4.350, train_loss=883.8950, val_loss=889.2964]
Training UniVI: 2%|▏ | 116/5000 [01:45<1:12:50, 1.12it/s, beta=1.250, gamma=4.350, train_loss=883.5805, val_loss=887.3410]
Training UniVI: 2%|▏ | 117/5000 [01:45<1:12:07, 1.13it/s, beta=1.250, gamma=4.350, train_loss=883.5805, val_loss=887.3410]
Training UniVI: 2%|▏ | 117/5000 [01:46<1:12:07, 1.13it/s, beta=1.250, gamma=4.350, train_loss=884.9154, val_loss=887.6201]
Training UniVI: 2%|▏ | 118/5000 [01:46<1:12:09, 1.13it/s, beta=1.250, gamma=4.350, train_loss=884.9154, val_loss=887.6201]
Training UniVI: 2%|▏ | 118/5000 [01:47<1:12:09, 1.13it/s, beta=1.250, gamma=4.350, train_loss=883.4442, val_loss=888.9057]
Training UniVI: 2%|▏ | 119/5000 [01:47<1:12:01, 1.13it/s, beta=1.250, gamma=4.350, train_loss=883.4442, val_loss=888.9057]
Training UniVI: 2%|▏ | 119/5000 [01:48<1:12:01, 1.13it/s, beta=1.250, gamma=4.350, train_loss=883.5108, val_loss=888.0222]
Training UniVI: 2%|▏ | 120/5000 [01:48<1:11:19, 1.14it/s, beta=1.250, gamma=4.350, train_loss=883.5108, val_loss=888.0222]
Training UniVI: 2%|▏ | 120/5000 [01:49<1:11:19, 1.14it/s, beta=1.250, gamma=4.350, train_loss=883.5896, val_loss=886.9872]
[2026-09-21 14:53:40,596] [UniVITrainer] [INFO] [Epoch 121] New best val loss: 886.9872
Training UniVI: 2%|▏ | 121/5000 [01:49<1:11:03, 1.14it/s, beta=1.250, gamma=4.350, train_loss=883.5896, val_loss=886.9872]
Training UniVI: 2%|▏ | 121/5000 [01:50<1:11:03, 1.14it/s, beta=1.250, gamma=4.350, train_loss=882.2176, val_loss=887.4818]
Training UniVI: 2%|▏ | 122/5000 [01:50<1:11:47, 1.13it/s, beta=1.250, gamma=4.350, train_loss=882.2176, val_loss=887.4818]
Training UniVI: 2%|▏ | 122/5000 [01:50<1:11:47, 1.13it/s, beta=1.250, gamma=4.350, train_loss=882.1806, val_loss=890.0005]
Training UniVI: 2%|▏ | 123/5000 [01:50<1:11:30, 1.14it/s, beta=1.250, gamma=4.350, train_loss=882.1806, val_loss=890.0005]
Training UniVI: 2%|▏ | 123/5000 [01:51<1:11:30, 1.14it/s, beta=1.250, gamma=4.350, train_loss=881.8607, val_loss=888.9534]
Training UniVI: 2%|▏ | 124/5000 [01:51<1:12:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=881.8607, val_loss=888.9534]
Training UniVI: 2%|▏ | 124/5000 [01:52<1:12:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=882.2139, val_loss=888.6741]
Training UniVI: 2%|▎ | 125/5000 [01:52<1:12:00, 1.13it/s, beta=1.250, gamma=4.350, train_loss=882.2139, val_loss=888.6741]
Training UniVI: 2%|▎ | 125/5000 [01:53<1:12:00, 1.13it/s, beta=1.250, gamma=4.350, train_loss=881.9089, val_loss=887.4033]
Training UniVI: 3%|▎ | 126/5000 [01:53<1:12:39, 1.12it/s, beta=1.250, gamma=4.350, train_loss=881.9089, val_loss=887.4033]
Training UniVI: 3%|▎ | 126/5000 [01:54<1:12:39, 1.12it/s, beta=1.250, gamma=4.350, train_loss=881.1110, val_loss=888.3115]
Training UniVI: 3%|▎ | 127/5000 [01:54<1:14:27, 1.09it/s, beta=1.250, gamma=4.350, train_loss=881.1110, val_loss=888.3115]
Training UniVI: 3%|▎ | 127/5000 [01:55<1:14:27, 1.09it/s, beta=1.250, gamma=4.350, train_loss=881.3662, val_loss=888.0859]
Training UniVI: 3%|▎ | 128/5000 [01:55<1:13:31, 1.10it/s, beta=1.250, gamma=4.350, train_loss=881.3662, val_loss=888.0859]
Training UniVI: 3%|▎ | 128/5000 [01:56<1:13:31, 1.10it/s, beta=1.250, gamma=4.350, train_loss=879.7921, val_loss=888.1489]
Training UniVI: 3%|▎ | 129/5000 [01:56<1:12:21, 1.12it/s, beta=1.250, gamma=4.350, train_loss=879.7921, val_loss=888.1489]
Training UniVI: 3%|▎ | 129/5000 [01:57<1:12:21, 1.12it/s, beta=1.250, gamma=4.350, train_loss=879.7201, val_loss=890.7105]
Training UniVI: 3%|▎ | 130/5000 [01:57<1:11:05, 1.14it/s, beta=1.250, gamma=4.350, train_loss=879.7201, val_loss=890.7105]
Training UniVI: 3%|▎ | 130/5000 [01:58<1:11:05, 1.14it/s, beta=1.250, gamma=4.350, train_loss=880.0296, val_loss=888.2458]
Training UniVI: 3%|▎ | 131/5000 [01:58<1:13:25, 1.11it/s, beta=1.250, gamma=4.350, train_loss=880.0296, val_loss=888.2458]
Training UniVI: 3%|▎ | 131/5000 [01:58<1:13:25, 1.11it/s, beta=1.250, gamma=4.350, train_loss=880.7845, val_loss=888.5060]
Training UniVI: 3%|▎ | 132/5000 [01:58<1:12:50, 1.11it/s, beta=1.250, gamma=4.350, train_loss=880.7845, val_loss=888.5060]
Training UniVI: 3%|▎ | 132/5000 [01:59<1:12:50, 1.11it/s, beta=1.250, gamma=4.350, train_loss=879.5491, val_loss=888.4536]
Training UniVI: 3%|▎ | 133/5000 [01:59<1:11:44, 1.13it/s, beta=1.250, gamma=4.350, train_loss=879.5491, val_loss=888.4536]
Training UniVI: 3%|▎ | 133/5000 [02:00<1:11:44, 1.13it/s, beta=1.250, gamma=4.350, train_loss=878.4768, val_loss=889.4156]
Training UniVI: 3%|▎ | 134/5000 [02:00<1:14:58, 1.08it/s, beta=1.250, gamma=4.350, train_loss=878.4768, val_loss=889.4156]
Training UniVI: 3%|▎ | 134/5000 [02:01<1:14:58, 1.08it/s, beta=1.250, gamma=4.350, train_loss=880.2407, val_loss=889.0700]
Training UniVI: 3%|▎ | 135/5000 [02:01<1:13:56, 1.10it/s, beta=1.250, gamma=4.350, train_loss=880.2407, val_loss=889.0700]
Training UniVI: 3%|▎ | 135/5000 [02:02<1:13:56, 1.10it/s, beta=1.250, gamma=4.350, train_loss=879.5400, val_loss=887.5726]
Training UniVI: 3%|▎ | 136/5000 [02:02<1:11:51, 1.13it/s, beta=1.250, gamma=4.350, train_loss=879.5400, val_loss=887.5726]
Training UniVI: 3%|▎ | 136/5000 [02:03<1:11:51, 1.13it/s, beta=1.250, gamma=4.350, train_loss=878.6791, val_loss=888.3956]
Training UniVI: 3%|▎ | 137/5000 [02:03<1:12:35, 1.12it/s, beta=1.250, gamma=4.350, train_loss=878.6791, val_loss=888.3956]
Training UniVI: 3%|▎ | 137/5000 [02:04<1:12:35, 1.12it/s, beta=1.250, gamma=4.350, train_loss=877.5706, val_loss=889.4424]
Training UniVI: 3%|▎ | 138/5000 [02:04<1:11:55, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.5706, val_loss=889.4424]
Training UniVI: 3%|▎ | 138/5000 [02:05<1:11:55, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.9592, val_loss=888.5991]
Training UniVI: 3%|▎ | 139/5000 [02:05<1:11:50, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.9592, val_loss=888.5991]
Training UniVI: 3%|▎ | 139/5000 [02:06<1:11:50, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.5908, val_loss=888.6234]
Training UniVI: 3%|▎ | 140/5000 [02:06<1:12:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=877.5908, val_loss=888.6234]
Training UniVI: 3%|▎ | 140/5000 [02:07<1:12:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=878.7180, val_loss=887.8015]
Training UniVI: 3%|▎ | 141/5000 [02:07<1:13:31, 1.10it/s, beta=1.250, gamma=4.350, train_loss=878.7180, val_loss=887.8015]
Training UniVI: 3%|▎ | 141/5000 [02:07<1:13:31, 1.10it/s, beta=1.250, gamma=4.350, train_loss=876.6455, val_loss=889.2848]
Training UniVI: 3%|▎ | 142/5000 [02:07<1:12:31, 1.12it/s, beta=1.250, gamma=4.350, train_loss=876.6455, val_loss=889.2848]
Training UniVI: 3%|▎ | 142/5000 [02:08<1:12:31, 1.12it/s, beta=1.250, gamma=4.350, train_loss=877.2353, val_loss=888.7197]
Training UniVI: 3%|▎ | 143/5000 [02:08<1:11:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.2353, val_loss=888.7197]
Training UniVI: 3%|▎ | 143/5000 [02:09<1:11:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=877.2689, val_loss=888.6655]
Training UniVI: 3%|▎ | 144/5000 [02:09<1:11:56, 1.12it/s, beta=1.250, gamma=4.350, train_loss=877.2689, val_loss=888.6655]
Training UniVI: 3%|▎ | 144/5000 [02:10<1:11:56, 1.12it/s, beta=1.250, gamma=4.350, train_loss=877.2432, val_loss=888.0015]
Training UniVI: 3%|▎ | 145/5000 [02:10<1:11:09, 1.14it/s, beta=1.250, gamma=4.350, train_loss=877.2432, val_loss=888.0015]
Training UniVI: 3%|▎ | 145/5000 [02:11<1:11:09, 1.14it/s, beta=1.250, gamma=4.350, train_loss=878.0694, val_loss=888.8857]
Training UniVI: 3%|▎ | 146/5000 [02:11<1:11:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=878.0694, val_loss=888.8857]
Training UniVI: 3%|▎ | 146/5000 [02:12<1:11:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=880.2223, val_loss=888.8651]
Training UniVI: 3%|▎ | 147/5000 [02:12<1:11:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=880.2223, val_loss=888.8651]
Training UniVI: 3%|▎ | 147/5000 [02:13<1:11:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=876.6502, val_loss=886.8277]
[2026-09-21 14:54:04,787] [UniVITrainer] [INFO] [Epoch 148] New best val loss: 886.8277
Training UniVI: 3%|▎ | 148/5000 [02:13<1:13:09, 1.11it/s, beta=1.250, gamma=4.350, train_loss=876.6502, val_loss=886.8277]
Training UniVI: 3%|▎ | 148/5000 [02:14<1:13:09, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.7525, val_loss=888.2861]
Training UniVI: 3%|▎ | 149/5000 [02:14<1:13:26, 1.10it/s, beta=1.250, gamma=4.350, train_loss=875.7525, val_loss=888.2861]
Training UniVI: 3%|▎ | 149/5000 [02:15<1:13:26, 1.10it/s, beta=1.250, gamma=4.350, train_loss=875.8846, val_loss=889.9709]
Training UniVI: 3%|▎ | 150/5000 [02:15<1:12:45, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.8846, val_loss=889.9709]
Training UniVI: 3%|▎ | 150/5000 [02:15<1:12:45, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.6537, val_loss=889.0258]
Training UniVI: 3%|▎ | 151/5000 [02:15<1:12:55, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.6537, val_loss=889.0258]
Training UniVI: 3%|▎ | 151/5000 [02:16<1:12:55, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.4267, val_loss=888.6857]
Training UniVI: 3%|▎ | 152/5000 [02:16<1:11:53, 1.12it/s, beta=1.250, gamma=4.350, train_loss=875.4267, val_loss=888.6857]
Training UniVI: 3%|▎ | 152/5000 [02:17<1:11:53, 1.12it/s, beta=1.250, gamma=4.350, train_loss=873.9799, val_loss=889.3954]
Training UniVI: 3%|▎ | 153/5000 [02:17<1:12:51, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.9799, val_loss=889.3954]
Training UniVI: 3%|▎ | 153/5000 [02:18<1:12:51, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.2254, val_loss=888.2807]
Training UniVI: 3%|▎ | 154/5000 [02:18<1:12:23, 1.12it/s, beta=1.250, gamma=4.350, train_loss=875.2254, val_loss=888.2807]
Training UniVI: 3%|▎ | 154/5000 [02:19<1:12:23, 1.12it/s, beta=1.250, gamma=4.350, train_loss=874.6284, val_loss=889.2197]
Training UniVI: 3%|▎ | 155/5000 [02:19<1:12:59, 1.11it/s, beta=1.250, gamma=4.350, train_loss=874.6284, val_loss=889.2197]
Training UniVI: 3%|▎ | 155/5000 [02:20<1:12:59, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.6917, val_loss=889.1626]
Training UniVI: 3%|▎ | 156/5000 [02:20<1:13:34, 1.10it/s, beta=1.250, gamma=4.350, train_loss=875.6917, val_loss=889.1626]
Training UniVI: 3%|▎ | 156/5000 [02:21<1:13:34, 1.10it/s, beta=1.250, gamma=4.350, train_loss=875.1629, val_loss=891.4322]
Training UniVI: 3%|▎ | 157/5000 [02:21<1:12:51, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.1629, val_loss=891.4322]
Training UniVI: 3%|▎ | 157/5000 [02:22<1:12:51, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.1331, val_loss=890.6423]
Training UniVI: 3%|▎ | 158/5000 [02:22<1:12:43, 1.11it/s, beta=1.250, gamma=4.350, train_loss=875.1331, val_loss=890.6423]
Training UniVI: 3%|▎ | 158/5000 [02:23<1:12:43, 1.11it/s, beta=1.250, gamma=4.350, train_loss=874.3728, val_loss=888.6396]
Training UniVI: 3%|▎ | 159/5000 [02:23<1:13:51, 1.09it/s, beta=1.250, gamma=4.350, train_loss=874.3728, val_loss=888.6396]
Training UniVI: 3%|▎ | 159/5000 [02:24<1:13:51, 1.09it/s, beta=1.250, gamma=4.350, train_loss=874.6416, val_loss=889.7513]
Training UniVI: 3%|▎ | 160/5000 [02:24<1:13:47, 1.09it/s, beta=1.250, gamma=4.350, train_loss=874.6416, val_loss=889.7513]
Training UniVI: 3%|▎ | 160/5000 [02:25<1:13:47, 1.09it/s, beta=1.250, gamma=4.350, train_loss=873.5157, val_loss=888.7291]
Training UniVI: 3%|▎ | 161/5000 [02:25<1:12:56, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.5157, val_loss=888.7291]
Training UniVI: 3%|▎ | 161/5000 [02:25<1:12:56, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.8598, val_loss=889.5291]
Training UniVI: 3%|▎ | 162/5000 [02:25<1:12:53, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.8598, val_loss=889.5291]
Training UniVI: 3%|▎ | 162/5000 [02:26<1:12:53, 1.11it/s, beta=1.250, gamma=4.350, train_loss=873.4055, val_loss=889.2480]
Training UniVI: 3%|▎ | 163/5000 [02:26<1:11:37, 1.13it/s, beta=1.250, gamma=4.350, train_loss=873.4055, val_loss=889.2480]
Training UniVI: 3%|▎ | 163/5000 [02:27<1:11:37, 1.13it/s, beta=1.250, gamma=4.350, train_loss=872.6863, val_loss=889.0416]
Training UniVI: 3%|▎ | 164/5000 [02:27<1:11:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=872.6863, val_loss=889.0416]
Training UniVI: 3%|▎ | 164/5000 [02:28<1:11:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=872.8122, val_loss=889.7844]
Training UniVI: 3%|▎ | 165/5000 [02:28<1:12:54, 1.11it/s, beta=1.250, gamma=4.350, train_loss=872.8122, val_loss=889.7844]
Training UniVI: 3%|▎ | 165/5000 [02:29<1:12:54, 1.11it/s, beta=1.250, gamma=4.350, train_loss=872.1457, val_loss=889.1287]
Training UniVI: 3%|▎ | 166/5000 [02:29<1:13:49, 1.09it/s, beta=1.250, gamma=4.350, train_loss=872.1457, val_loss=889.1287]
Training UniVI: 3%|▎ | 166/5000 [02:30<1:13:49, 1.09it/s, beta=1.250, gamma=4.350, train_loss=872.4584, val_loss=889.9017]
Training UniVI: 3%|▎ | 167/5000 [02:30<1:14:26, 1.08it/s, beta=1.250, gamma=4.350, train_loss=872.4584, val_loss=889.9017]
Training UniVI: 3%|▎ | 167/5000 [02:31<1:14:26, 1.08it/s, beta=1.250, gamma=4.350, train_loss=870.7746, val_loss=890.3396]
Training UniVI: 3%|▎ | 168/5000 [02:31<1:12:27, 1.11it/s, beta=1.250, gamma=4.350, train_loss=870.7746, val_loss=890.3396]
Training UniVI: 3%|▎ | 168/5000 [02:32<1:12:27, 1.11it/s, beta=1.250, gamma=4.350, train_loss=872.0742, val_loss=890.3478]
Training UniVI: 3%|▎ | 169/5000 [02:32<1:12:00, 1.12it/s, beta=1.250, gamma=4.350, train_loss=872.0742, val_loss=890.3478]
Training UniVI: 3%|▎ | 169/5000 [02:33<1:12:00, 1.12it/s, beta=1.250, gamma=4.350, train_loss=871.9232, val_loss=890.9636]
Training UniVI: 3%|▎ | 170/5000 [02:33<1:10:42, 1.14it/s, beta=1.250, gamma=4.350, train_loss=871.9232, val_loss=890.9636]
Training UniVI: 3%|▎ | 170/5000 [02:33<1:10:42, 1.14it/s, beta=1.250, gamma=4.350, train_loss=871.9795, val_loss=889.0049]
Training UniVI: 3%|▎ | 171/5000 [02:33<1:09:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=871.9795, val_loss=889.0049]
Training UniVI: 3%|▎ | 171/5000 [02:34<1:09:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=872.5932, val_loss=889.4550]
Training UniVI: 3%|▎ | 172/5000 [02:34<1:09:55, 1.15it/s, beta=1.250, gamma=4.350, train_loss=872.5932, val_loss=889.4550]
Training UniVI: 3%|▎ | 172/5000 [02:35<1:09:55, 1.15it/s, beta=1.250, gamma=4.350, train_loss=871.0508, val_loss=890.7028]
Training UniVI: 3%|▎ | 173/5000 [02:35<1:09:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=871.0508, val_loss=890.7028]
Training UniVI: 3%|▎ | 173/5000 [02:36<1:09:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=870.8006, val_loss=889.3295]
Training UniVI: 3%|▎ | 174/5000 [02:36<1:14:42, 1.08it/s, beta=1.250, gamma=4.350, train_loss=870.8006, val_loss=889.3295]
Training UniVI: 3%|▎ | 174/5000 [02:37<1:14:42, 1.08it/s, beta=1.250, gamma=4.350, train_loss=870.1096, val_loss=889.0342]
Training UniVI: 4%|▎ | 175/5000 [02:37<1:14:55, 1.07it/s, beta=1.250, gamma=4.350, train_loss=870.1096, val_loss=889.0342]
Training UniVI: 4%|▎ | 175/5000 [02:38<1:14:55, 1.07it/s, beta=1.250, gamma=4.350, train_loss=870.9928, val_loss=889.4808]
Training UniVI: 4%|▎ | 176/5000 [02:38<1:13:20, 1.10it/s, beta=1.250, gamma=4.350, train_loss=870.9928, val_loss=889.4808]
Training UniVI: 4%|▎ | 176/5000 [02:39<1:13:20, 1.10it/s, beta=1.250, gamma=4.350, train_loss=870.1784, val_loss=889.5366]
Training UniVI: 4%|▎ | 177/5000 [02:39<1:13:24, 1.10it/s, beta=1.250, gamma=4.350, train_loss=870.1784, val_loss=889.5366]
Training UniVI: 4%|▎ | 177/5000 [02:40<1:13:24, 1.10it/s, beta=1.250, gamma=4.350, train_loss=870.9192, val_loss=890.7540]
Training UniVI: 4%|▎ | 178/5000 [02:40<1:12:18, 1.11it/s, beta=1.250, gamma=4.350, train_loss=870.9192, val_loss=890.7540]
Training UniVI: 4%|▎ | 178/5000 [02:41<1:12:18, 1.11it/s, beta=1.250, gamma=4.350, train_loss=868.3253, val_loss=891.0513]
Training UniVI: 4%|▎ | 179/5000 [02:41<1:11:54, 1.12it/s, beta=1.250, gamma=4.350, train_loss=868.3253, val_loss=891.0513]
Training UniVI: 4%|▎ | 179/5000 [02:42<1:11:54, 1.12it/s, beta=1.250, gamma=4.350, train_loss=869.2395, val_loss=888.3149]
Training UniVI: 4%|▎ | 180/5000 [02:42<1:11:15, 1.13it/s, beta=1.250, gamma=4.350, train_loss=869.2395, val_loss=888.3149]
Training UniVI: 4%|▎ | 180/5000 [02:42<1:11:15, 1.13it/s, beta=1.250, gamma=4.350, train_loss=870.3498, val_loss=891.4676]
Training UniVI: 4%|▎ | 181/5000 [02:42<1:11:08, 1.13it/s, beta=1.250, gamma=4.350, train_loss=870.3498, val_loss=891.4676]
Training UniVI: 4%|▎ | 181/5000 [02:43<1:11:08, 1.13it/s, beta=1.250, gamma=4.350, train_loss=868.5311, val_loss=889.8113]
Training UniVI: 4%|▎ | 182/5000 [02:43<1:11:56, 1.12it/s, beta=1.250, gamma=4.350, train_loss=868.5311, val_loss=889.8113]
Training UniVI: 4%|▎ | 182/5000 [02:44<1:11:56, 1.12it/s, beta=1.250, gamma=4.350, train_loss=869.6364, val_loss=892.1692]
Training UniVI: 4%|▎ | 183/5000 [02:44<1:11:04, 1.13it/s, beta=1.250, gamma=4.350, train_loss=869.6364, val_loss=892.1692]
Training UniVI: 4%|▎ | 183/5000 [02:45<1:11:04, 1.13it/s, beta=1.250, gamma=4.350, train_loss=869.5589, val_loss=890.2881]
Training UniVI: 4%|▎ | 184/5000 [02:45<1:11:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=869.5589, val_loss=890.2881]
Training UniVI: 4%|▎ | 184/5000 [02:46<1:11:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=868.2935, val_loss=890.7543]
Training UniVI: 4%|▎ | 185/5000 [02:46<1:11:22, 1.12it/s, beta=1.250, gamma=4.350, train_loss=868.2935, val_loss=890.7543]
Training UniVI: 4%|▎ | 185/5000 [02:47<1:11:22, 1.12it/s, beta=1.250, gamma=4.350, train_loss=868.4095, val_loss=889.8688]
Training UniVI: 4%|▎ | 186/5000 [02:47<1:10:05, 1.14it/s, beta=1.250, gamma=4.350, train_loss=868.4095, val_loss=889.8688]
Training UniVI: 4%|▎ | 186/5000 [02:48<1:10:05, 1.14it/s, beta=1.250, gamma=4.350, train_loss=870.1429, val_loss=888.8198]
Training UniVI: 4%|▎ | 187/5000 [02:48<1:10:16, 1.14it/s, beta=1.250, gamma=4.350, train_loss=870.1429, val_loss=888.8198]
Training UniVI: 4%|▎ | 187/5000 [02:49<1:10:16, 1.14it/s, beta=1.250, gamma=4.350, train_loss=868.7343, val_loss=889.8396]
Training UniVI: 4%|▍ | 188/5000 [02:49<1:09:39, 1.15it/s, beta=1.250, gamma=4.350, train_loss=868.7343, val_loss=889.8396]
Training UniVI: 4%|▍ | 188/5000 [02:49<1:09:39, 1.15it/s, beta=1.250, gamma=4.350, train_loss=868.5825, val_loss=890.8065]
Training UniVI: 4%|▍ | 189/5000 [02:49<1:08:57, 1.16it/s, beta=1.250, gamma=4.350, train_loss=868.5825, val_loss=890.8065]
Training UniVI: 4%|▍ | 189/5000 [02:50<1:08:57, 1.16it/s, beta=1.250, gamma=4.350, train_loss=868.9248, val_loss=891.3485]
Training UniVI: 4%|▍ | 190/5000 [02:50<1:09:53, 1.15it/s, beta=1.250, gamma=4.350, train_loss=868.9248, val_loss=891.3485]
Training UniVI: 4%|▍ | 190/5000 [02:51<1:09:53, 1.15it/s, beta=1.250, gamma=4.350, train_loss=869.2510, val_loss=890.9189]
Training UniVI: 4%|▍ | 191/5000 [02:51<1:10:14, 1.14it/s, beta=1.250, gamma=4.350, train_loss=869.2510, val_loss=890.9189]
Training UniVI: 4%|▍ | 191/5000 [02:52<1:10:14, 1.14it/s, beta=1.250, gamma=4.350, train_loss=869.0915, val_loss=891.0560]
Training UniVI: 4%|▍ | 192/5000 [02:52<1:11:07, 1.13it/s, beta=1.250, gamma=4.350, train_loss=869.0915, val_loss=891.0560]
Training UniVI: 4%|▍ | 192/5000 [02:53<1:11:07, 1.13it/s, beta=1.250, gamma=4.350, train_loss=868.9463, val_loss=891.7064]
Training UniVI: 4%|▍ | 193/5000 [02:53<1:10:27, 1.14it/s, beta=1.250, gamma=4.350, train_loss=868.9463, val_loss=891.7064]
Training UniVI: 4%|▍ | 193/5000 [02:54<1:10:27, 1.14it/s, beta=1.250, gamma=4.350, train_loss=867.9920, val_loss=890.2773]
Training UniVI: 4%|▍ | 194/5000 [02:54<1:11:30, 1.12it/s, beta=1.250, gamma=4.350, train_loss=867.9920, val_loss=890.2773]
Training UniVI: 4%|▍ | 194/5000 [02:55<1:11:30, 1.12it/s, beta=1.250, gamma=4.350, train_loss=866.2998, val_loss=891.4692]
Training UniVI: 4%|▍ | 195/5000 [02:55<1:11:01, 1.13it/s, beta=1.250, gamma=4.350, train_loss=866.2998, val_loss=891.4692]
Training UniVI: 4%|▍ | 195/5000 [02:56<1:11:01, 1.13it/s, beta=1.250, gamma=4.350, train_loss=868.0017, val_loss=890.7931]
Training UniVI: 4%|▍ | 196/5000 [02:56<1:10:18, 1.14it/s, beta=1.250, gamma=4.350, train_loss=868.0017, val_loss=890.7931]
Training UniVI: 4%|▍ | 196/5000 [02:57<1:10:18, 1.14it/s, beta=1.250, gamma=4.350, train_loss=867.2565, val_loss=889.4823]
Training UniVI: 4%|▍ | 197/5000 [02:57<1:10:24, 1.14it/s, beta=1.250, gamma=4.350, train_loss=867.2565, val_loss=889.4823]
Training UniVI: 4%|▍ | 197/5000 [02:57<1:10:24, 1.14it/s, beta=1.250, gamma=4.350, train_loss=866.3946, val_loss=890.8890]
Training UniVI: 4%|▍ | 198/5000 [02:57<1:09:15, 1.16it/s, beta=1.250, gamma=4.350, train_loss=866.3946, val_loss=890.8890]
Training UniVI: 4%|▍ | 198/5000 [02:58<1:09:15, 1.16it/s, beta=1.250, gamma=4.350, train_loss=866.4987, val_loss=889.5818]
Training UniVI: 4%|▍ | 199/5000 [02:58<1:10:02, 1.14it/s, beta=1.250, gamma=4.350, train_loss=866.4987, val_loss=889.5818]
[2026-09-21 14:54:51,075] [UniVITrainer] [INFO] [Epoch 200] Train loss=866.7147 (beta=1.250, gamma=4.350)
[2026-09-21 14:54:51,099] [UniVITrainer] [INFO] [Epoch 200] Val loss=889.2207 (beta=1.250, gamma=4.350)
Training UniVI: 4%|▍ | 199/5000 [02:59<1:10:02, 1.14it/s, beta=1.250, gamma=4.350, train_loss=866.7147, val_loss=889.2207]
Training UniVI: 4%|▍ | 200/5000 [02:59<1:09:20, 1.15it/s, beta=1.250, gamma=4.350, train_loss=866.7147, val_loss=889.2207]
Training UniVI: 4%|▍ | 200/5000 [03:00<1:09:20, 1.15it/s, beta=1.250, gamma=4.350, train_loss=868.3069, val_loss=891.1461]
Training UniVI: 4%|▍ | 201/5000 [03:00<1:10:38, 1.13it/s, beta=1.250, gamma=4.350, train_loss=868.3069, val_loss=891.1461]
Training UniVI: 4%|▍ | 201/5000 [03:01<1:10:38, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.6701, val_loss=891.8705]
Training UniVI: 4%|▍ | 202/5000 [03:01<1:10:42, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.6701, val_loss=891.8705]
Training UniVI: 4%|▍ | 202/5000 [03:02<1:10:42, 1.13it/s, beta=1.250, gamma=4.350, train_loss=866.3322, val_loss=890.2728]
Training UniVI: 4%|▍ | 203/5000 [03:02<1:09:54, 1.14it/s, beta=1.250, gamma=4.350, train_loss=866.3322, val_loss=890.2728]
Training UniVI: 4%|▍ | 203/5000 [03:03<1:09:54, 1.14it/s, beta=1.250, gamma=4.350, train_loss=864.8278, val_loss=890.4873]
Training UniVI: 4%|▍ | 204/5000 [03:03<1:10:37, 1.13it/s, beta=1.250, gamma=4.350, train_loss=864.8278, val_loss=890.4873]
Training UniVI: 4%|▍ | 204/5000 [03:04<1:10:37, 1.13it/s, beta=1.250, gamma=4.350, train_loss=866.4613, val_loss=890.4841]
Training UniVI: 4%|▍ | 205/5000 [03:04<1:10:27, 1.13it/s, beta=1.250, gamma=4.350, train_loss=866.4613, val_loss=890.4841]
Training UniVI: 4%|▍ | 205/5000 [03:04<1:10:27, 1.13it/s, beta=1.250, gamma=4.350, train_loss=864.7845, val_loss=891.3320]
Training UniVI: 4%|▍ | 206/5000 [03:04<1:11:11, 1.12it/s, beta=1.250, gamma=4.350, train_loss=864.7845, val_loss=891.3320]
Training UniVI: 4%|▍ | 206/5000 [03:05<1:11:11, 1.12it/s, beta=1.250, gamma=4.350, train_loss=865.4617, val_loss=889.9847]
Training UniVI: 4%|▍ | 207/5000 [03:05<1:10:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.4617, val_loss=889.9847]
Training UniVI: 4%|▍ | 207/5000 [03:06<1:10:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=864.0515, val_loss=889.2097]
Training UniVI: 4%|▍ | 208/5000 [03:06<1:13:43, 1.08it/s, beta=1.250, gamma=4.350, train_loss=864.0515, val_loss=889.2097]
Training UniVI: 4%|▍ | 208/5000 [03:07<1:13:43, 1.08it/s, beta=1.250, gamma=4.350, train_loss=865.9423, val_loss=890.5394]
Training UniVI: 4%|▍ | 209/5000 [03:07<1:12:02, 1.11it/s, beta=1.250, gamma=4.350, train_loss=865.9423, val_loss=890.5394]
Training UniVI: 4%|▍ | 209/5000 [03:08<1:12:02, 1.11it/s, beta=1.250, gamma=4.350, train_loss=863.9802, val_loss=888.9445]
Training UniVI: 4%|▍ | 210/5000 [03:08<1:11:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=863.9802, val_loss=888.9445]
Training UniVI: 4%|▍ | 210/5000 [03:09<1:11:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=863.8595, val_loss=888.7031]
Training UniVI: 4%|▍ | 211/5000 [03:09<1:10:25, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.8595, val_loss=888.7031]
Training UniVI: 4%|▍ | 211/5000 [03:10<1:10:25, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.2064, val_loss=892.6105]
Training UniVI: 4%|▍ | 212/5000 [03:10<1:10:08, 1.14it/s, beta=1.250, gamma=4.350, train_loss=865.2064, val_loss=892.6105]
Training UniVI: 4%|▍ | 212/5000 [03:11<1:10:08, 1.14it/s, beta=1.250, gamma=4.350, train_loss=864.9552, val_loss=890.8774]
Training UniVI: 4%|▍ | 213/5000 [03:11<1:09:11, 1.15it/s, beta=1.250, gamma=4.350, train_loss=864.9552, val_loss=890.8774]
Training UniVI: 4%|▍ | 213/5000 [03:11<1:09:11, 1.15it/s, beta=1.250, gamma=4.350, train_loss=865.7500, val_loss=891.1387]
Training UniVI: 4%|▍ | 214/5000 [03:11<1:08:57, 1.16it/s, beta=1.250, gamma=4.350, train_loss=865.7500, val_loss=891.1387]
Training UniVI: 4%|▍ | 214/5000 [03:12<1:08:57, 1.16it/s, beta=1.250, gamma=4.350, train_loss=864.7678, val_loss=892.1287]
Training UniVI: 4%|▍ | 215/5000 [03:12<1:10:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=864.7678, val_loss=892.1287]
Training UniVI: 4%|▍ | 215/5000 [03:13<1:10:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.6777, val_loss=889.1742]
Training UniVI: 4%|▍ | 216/5000 [03:13<1:10:35, 1.13it/s, beta=1.250, gamma=4.350, train_loss=865.6777, val_loss=889.1742]
Training UniVI: 4%|▍ | 216/5000 [03:14<1:10:35, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.5862, val_loss=891.0144]
Training UniVI: 4%|▍ | 217/5000 [03:14<1:11:29, 1.11it/s, beta=1.250, gamma=4.350, train_loss=863.5862, val_loss=891.0144]
Training UniVI: 4%|▍ | 217/5000 [03:15<1:11:29, 1.11it/s, beta=1.250, gamma=4.350, train_loss=863.3299, val_loss=890.6094]
Training UniVI: 4%|▍ | 218/5000 [03:15<1:12:10, 1.10it/s, beta=1.250, gamma=4.350, train_loss=863.3299, val_loss=890.6094]
Training UniVI: 4%|▍ | 218/5000 [03:16<1:12:10, 1.10it/s, beta=1.250, gamma=4.350, train_loss=864.2755, val_loss=892.1498]
Training UniVI: 4%|▍ | 219/5000 [03:16<1:11:34, 1.11it/s, beta=1.250, gamma=4.350, train_loss=864.2755, val_loss=892.1498]
Training UniVI: 4%|▍ | 219/5000 [03:17<1:11:34, 1.11it/s, beta=1.250, gamma=4.350, train_loss=862.6412, val_loss=890.8795]
Training UniVI: 4%|▍ | 220/5000 [03:17<1:11:04, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.6412, val_loss=890.8795]
Training UniVI: 4%|▍ | 220/5000 [03:18<1:11:04, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.1179, val_loss=890.1318]
Training UniVI: 4%|▍ | 221/5000 [03:18<1:10:29, 1.13it/s, beta=1.250, gamma=4.350, train_loss=862.1179, val_loss=890.1318]
Training UniVI: 4%|▍ | 221/5000 [03:19<1:10:29, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.2473, val_loss=891.0784]
Training UniVI: 4%|▍ | 222/5000 [03:19<1:10:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.2473, val_loss=891.0784]
Training UniVI: 4%|▍ | 222/5000 [03:20<1:10:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.0424, val_loss=891.2137]
Training UniVI: 4%|▍ | 223/5000 [03:20<1:11:37, 1.11it/s, beta=1.250, gamma=4.350, train_loss=863.0424, val_loss=891.2137]
Training UniVI: 4%|▍ | 223/5000 [03:20<1:11:37, 1.11it/s, beta=1.250, gamma=4.350, train_loss=863.0003, val_loss=891.1519]
Training UniVI: 4%|▍ | 224/5000 [03:20<1:10:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.0003, val_loss=891.1519]
Training UniVI: 4%|▍ | 224/5000 [03:21<1:10:31, 1.13it/s, beta=1.250, gamma=4.350, train_loss=863.1082, val_loss=891.6797]
Training UniVI: 4%|▍ | 225/5000 [03:21<1:09:36, 1.14it/s, beta=1.250, gamma=4.350, train_loss=863.1082, val_loss=891.6797]
Training UniVI: 4%|▍ | 225/5000 [03:22<1:09:36, 1.14it/s, beta=1.250, gamma=4.350, train_loss=862.9762, val_loss=892.1388]
Training UniVI: 5%|▍ | 226/5000 [03:22<1:10:15, 1.13it/s, beta=1.250, gamma=4.350, train_loss=862.9762, val_loss=892.1388]
Training UniVI: 5%|▍ | 226/5000 [03:23<1:10:15, 1.13it/s, beta=1.250, gamma=4.350, train_loss=862.9105, val_loss=891.6038]
Training UniVI: 5%|▍ | 227/5000 [03:23<1:10:36, 1.13it/s, beta=1.250, gamma=4.350, train_loss=862.9105, val_loss=891.6038]
Training UniVI: 5%|▍ | 227/5000 [03:24<1:10:36, 1.13it/s, beta=1.250, gamma=4.350, train_loss=861.6343, val_loss=892.8597]
Training UniVI: 5%|▍ | 228/5000 [03:24<1:11:27, 1.11it/s, beta=1.250, gamma=4.350, train_loss=861.6343, val_loss=892.8597]
Training UniVI: 5%|▍ | 228/5000 [03:25<1:11:27, 1.11it/s, beta=1.250, gamma=4.350, train_loss=860.6316, val_loss=890.8265]
Training UniVI: 5%|▍ | 229/5000 [03:25<1:11:04, 1.12it/s, beta=1.250, gamma=4.350, train_loss=860.6316, val_loss=890.8265]
Training UniVI: 5%|▍ | 229/5000 [03:26<1:11:04, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.4595, val_loss=890.6915]
Training UniVI: 5%|▍ | 230/5000 [03:26<1:10:47, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.4595, val_loss=890.6915]
Training UniVI: 5%|▍ | 230/5000 [03:27<1:10:47, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.0265, val_loss=892.0892]
Training UniVI: 5%|▍ | 231/5000 [03:27<1:10:32, 1.13it/s, beta=1.250, gamma=4.350, train_loss=862.0265, val_loss=892.0892]
Training UniVI: 5%|▍ | 231/5000 [03:28<1:10:32, 1.13it/s, beta=1.250, gamma=4.350, train_loss=861.7953, val_loss=892.4904]
Training UniVI: 5%|▍ | 232/5000 [03:28<1:10:48, 1.12it/s, beta=1.250, gamma=4.350, train_loss=861.7953, val_loss=892.4904]
Training UniVI: 5%|▍ | 232/5000 [03:28<1:10:48, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.6305, val_loss=891.8813]
Training UniVI: 5%|▍ | 233/5000 [03:28<1:11:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=862.6305, val_loss=891.8813]
Training UniVI: 5%|▍ | 233/5000 [03:29<1:11:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=861.7685, val_loss=890.7217]
Training UniVI: 5%|▍ | 234/5000 [03:29<1:09:52, 1.14it/s, beta=1.250, gamma=4.350, train_loss=861.7685, val_loss=890.7217]
Training UniVI: 5%|▍ | 234/5000 [03:30<1:09:52, 1.14it/s, beta=1.250, gamma=4.350, train_loss=861.1605, val_loss=892.2935]
Training UniVI: 5%|▍ | 235/5000 [03:30<1:10:02, 1.13it/s, beta=1.250, gamma=4.350, train_loss=861.1605, val_loss=892.2935]
Training UniVI: 5%|▍ | 235/5000 [03:31<1:10:02, 1.13it/s, beta=1.250, gamma=4.350, train_loss=860.0218, val_loss=892.5731]
Training UniVI: 5%|▍ | 236/5000 [03:31<1:09:15, 1.15it/s, beta=1.250, gamma=4.350, train_loss=860.0218, val_loss=892.5731]
Training UniVI: 5%|▍ | 236/5000 [03:32<1:09:15, 1.15it/s, beta=1.250, gamma=4.350, train_loss=860.1054, val_loss=891.3054]
Training UniVI: 5%|▍ | 237/5000 [03:32<1:10:36, 1.12it/s, beta=1.250, gamma=4.350, train_loss=860.1054, val_loss=891.3054]
Training UniVI: 5%|▍ | 237/5000 [03:33<1:10:36, 1.12it/s, beta=1.250, gamma=4.350, train_loss=860.7519, val_loss=892.0619]
Training UniVI: 5%|▍ | 238/5000 [03:33<1:09:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=860.7519, val_loss=892.0619]
Training UniVI: 5%|▍ | 238/5000 [03:34<1:09:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=860.7203, val_loss=892.7811]
Training UniVI: 5%|▍ | 239/5000 [03:34<1:09:05, 1.15it/s, beta=1.250, gamma=4.350, train_loss=860.7203, val_loss=892.7811]
Training UniVI: 5%|▍ | 239/5000 [03:35<1:09:05, 1.15it/s, beta=1.250, gamma=4.350, train_loss=859.4758, val_loss=892.7810]
Training UniVI: 5%|▍ | 240/5000 [03:35<1:09:17, 1.14it/s, beta=1.250, gamma=4.350, train_loss=859.4758, val_loss=892.7810]
Training UniVI: 5%|▍ | 240/5000 [03:35<1:09:17, 1.14it/s, beta=1.250, gamma=4.350, train_loss=861.0663, val_loss=893.8797]
Training UniVI: 5%|▍ | 241/5000 [03:35<1:09:09, 1.15it/s, beta=1.250, gamma=4.350, train_loss=861.0663, val_loss=893.8797]
Training UniVI: 5%|▍ | 241/5000 [03:36<1:09:09, 1.15it/s, beta=1.250, gamma=4.350, train_loss=861.0526, val_loss=894.0498]
Training UniVI: 5%|▍ | 242/5000 [03:36<1:12:58, 1.09it/s, beta=1.250, gamma=4.350, train_loss=861.0526, val_loss=894.0498]
Training UniVI: 5%|▍ | 242/5000 [03:37<1:12:58, 1.09it/s, beta=1.250, gamma=4.350, train_loss=861.5787, val_loss=893.6901]
Training UniVI: 5%|▍ | 243/5000 [03:37<1:12:02, 1.10it/s, beta=1.250, gamma=4.350, train_loss=861.5787, val_loss=893.6901]
Training UniVI: 5%|▍ | 243/5000 [03:38<1:12:02, 1.10it/s, beta=1.250, gamma=4.350, train_loss=860.4533, val_loss=892.4972]
Training UniVI: 5%|▍ | 244/5000 [03:38<1:12:42, 1.09it/s, beta=1.250, gamma=4.350, train_loss=860.4533, val_loss=892.4972]
Training UniVI: 5%|▍ | 244/5000 [03:39<1:12:42, 1.09it/s, beta=1.250, gamma=4.350, train_loss=859.6990, val_loss=892.9866]
Training UniVI: 5%|▍ | 245/5000 [03:39<1:11:39, 1.11it/s, beta=1.250, gamma=4.350, train_loss=859.6990, val_loss=892.9866]
Training UniVI: 5%|▍ | 245/5000 [03:40<1:11:39, 1.11it/s, beta=1.250, gamma=4.350, train_loss=859.1462, val_loss=894.0923]
Training UniVI: 5%|▍ | 246/5000 [03:40<1:11:52, 1.10it/s, beta=1.250, gamma=4.350, train_loss=859.1462, val_loss=894.0923]
Training UniVI: 5%|▍ | 246/5000 [03:41<1:11:52, 1.10it/s, beta=1.250, gamma=4.350, train_loss=858.9024, val_loss=894.1658]
Training UniVI: 5%|▍ | 247/5000 [03:41<1:10:54, 1.12it/s, beta=1.250, gamma=4.350, train_loss=858.9024, val_loss=894.1658]
Training UniVI: 5%|▍ | 247/5000 [03:42<1:10:54, 1.12it/s, beta=1.250, gamma=4.350, train_loss=859.2211, val_loss=891.9868]
Training UniVI: 5%|▍ | 248/5000 [03:42<1:10:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=859.2211, val_loss=891.9868]
Training UniVI: 5%|▍ | 248/5000 [03:43<1:10:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=858.3936, val_loss=894.3184]
Training UniVI: 5%|▍ | 249/5000 [03:43<1:10:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=858.3936, val_loss=894.3184]
Training UniVI: 5%|▍ | 249/5000 [03:44<1:10:19, 1.13it/s, beta=1.250, gamma=4.350, train_loss=858.3781, val_loss=893.5517]
Training UniVI: 5%|▌ | 250/5000 [03:44<1:09:43, 1.14it/s, beta=1.250, gamma=4.350, train_loss=858.3781, val_loss=893.5517]
Training UniVI: 5%|▌ | 250/5000 [03:44<1:09:43, 1.14it/s, beta=1.250, gamma=4.350, train_loss=857.4314, val_loss=893.1679]
Training UniVI: 5%|▌ | 251/5000 [03:44<1:10:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=857.4314, val_loss=893.1679]
Training UniVI: 5%|▌ | 251/5000 [03:45<1:10:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=858.6362, val_loss=892.8218]
Training UniVI: 5%|▌ | 252/5000 [03:45<1:09:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=858.6362, val_loss=892.8218]
Training UniVI: 5%|▌ | 252/5000 [03:46<1:09:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=857.2685, val_loss=894.7858]
Training UniVI: 5%|▌ | 253/5000 [03:46<1:09:53, 1.13it/s, beta=1.250, gamma=4.350, train_loss=857.2685, val_loss=894.7858]
Training UniVI: 5%|▌ | 253/5000 [03:47<1:09:53, 1.13it/s, beta=1.250, gamma=4.350, train_loss=859.7059, val_loss=893.6790]
Training UniVI: 5%|▌ | 254/5000 [03:47<1:09:11, 1.14it/s, beta=1.250, gamma=4.350, train_loss=859.7059, val_loss=893.6790]
Training UniVI: 5%|▌ | 254/5000 [03:48<1:09:11, 1.14it/s, beta=1.250, gamma=4.350, train_loss=860.5280, val_loss=891.9516]
Training UniVI: 5%|▌ | 255/5000 [03:48<1:08:45, 1.15it/s, beta=1.250, gamma=4.350, train_loss=860.5280, val_loss=891.9516]
Training UniVI: 5%|▌ | 255/5000 [03:49<1:08:45, 1.15it/s, beta=1.250, gamma=4.350, train_loss=857.7251, val_loss=891.3953]
Training UniVI: 5%|▌ | 256/5000 [03:49<1:09:23, 1.14it/s, beta=1.250, gamma=4.350, train_loss=857.7251, val_loss=891.3953]
Training UniVI: 5%|▌ | 256/5000 [03:50<1:09:23, 1.14it/s, beta=1.250, gamma=4.350, train_loss=856.8768, val_loss=893.1881]
Training UniVI: 5%|▌ | 257/5000 [03:50<1:08:18, 1.16it/s, beta=1.250, gamma=4.350, train_loss=856.8768, val_loss=893.1881]
Training UniVI: 5%|▌ | 257/5000 [03:51<1:08:18, 1.16it/s, beta=1.250, gamma=4.350, train_loss=857.6478, val_loss=892.8001]
Training UniVI: 5%|▌ | 258/5000 [03:51<1:08:48, 1.15it/s, beta=1.250, gamma=4.350, train_loss=857.6478, val_loss=892.8001]
Training UniVI: 5%|▌ | 258/5000 [03:51<1:08:48, 1.15it/s, beta=1.250, gamma=4.350, train_loss=858.8979, val_loss=893.8575]
Training UniVI: 5%|▌ | 259/5000 [03:51<1:07:53, 1.16it/s, beta=1.250, gamma=4.350, train_loss=858.8979, val_loss=893.8575]
Training UniVI: 5%|▌ | 259/5000 [03:52<1:07:53, 1.16it/s, beta=1.250, gamma=4.350, train_loss=858.5451, val_loss=894.0128]
Training UniVI: 5%|▌ | 260/5000 [03:52<1:09:19, 1.14it/s, beta=1.250, gamma=4.350, train_loss=858.5451, val_loss=894.0128]
Training UniVI: 5%|▌ | 260/5000 [03:53<1:09:19, 1.14it/s, beta=1.250, gamma=4.350, train_loss=857.9398, val_loss=893.0226]
Training UniVI: 5%|▌ | 261/5000 [03:53<1:08:35, 1.15it/s, beta=1.250, gamma=4.350, train_loss=857.9398, val_loss=893.0226]
Training UniVI: 5%|▌ | 261/5000 [03:54<1:08:35, 1.15it/s, beta=1.250, gamma=4.350, train_loss=856.6895, val_loss=890.7070]
Training UniVI: 5%|▌ | 262/5000 [03:54<1:10:21, 1.12it/s, beta=1.250, gamma=4.350, train_loss=856.6895, val_loss=890.7070]
Training UniVI: 5%|▌ | 262/5000 [03:55<1:10:21, 1.12it/s, beta=1.250, gamma=4.350, train_loss=856.8003, val_loss=894.0382]
Training UniVI: 5%|▌ | 263/5000 [03:55<1:21:10, 1.03s/it, beta=1.250, gamma=4.350, train_loss=856.8003, val_loss=894.0382]
Training UniVI: 5%|▌ | 263/5000 [03:57<1:21:10, 1.03s/it, beta=1.250, gamma=4.350, train_loss=856.4870, val_loss=895.3937]
Training UniVI: 5%|▌ | 264/5000 [03:57<1:30:58, 1.15s/it, beta=1.250, gamma=4.350, train_loss=856.4870, val_loss=895.3937]
Training UniVI: 5%|▌ | 264/5000 [03:58<1:30:58, 1.15s/it, beta=1.250, gamma=4.350, train_loss=857.0018, val_loss=892.0891]
Training UniVI: 5%|▌ | 265/5000 [03:58<1:31:40, 1.16s/it, beta=1.250, gamma=4.350, train_loss=857.0018, val_loss=892.0891]
Training UniVI: 5%|▌ | 265/5000 [03:59<1:31:40, 1.16s/it, beta=1.250, gamma=4.350, train_loss=855.8613, val_loss=894.6597]
Training UniVI: 5%|▌ | 266/5000 [03:59<1:26:19, 1.09s/it, beta=1.250, gamma=4.350, train_loss=855.8613, val_loss=894.6597]
Training UniVI: 5%|▌ | 266/5000 [04:00<1:26:19, 1.09s/it, beta=1.250, gamma=4.350, train_loss=855.5712, val_loss=895.0818]
Training UniVI: 5%|▌ | 267/5000 [04:00<1:22:13, 1.04s/it, beta=1.250, gamma=4.350, train_loss=855.5712, val_loss=895.0818]
Training UniVI: 5%|▌ | 267/5000 [04:01<1:22:13, 1.04s/it, beta=1.250, gamma=4.350, train_loss=857.4423, val_loss=893.4550]
Training UniVI: 5%|▌ | 268/5000 [04:01<1:18:21, 1.01it/s, beta=1.250, gamma=4.350, train_loss=857.4423, val_loss=893.4550]
Training UniVI: 5%|▌ | 268/5000 [04:02<1:18:21, 1.01it/s, beta=1.250, gamma=4.350, train_loss=856.9115, val_loss=892.8193]
Training UniVI: 5%|▌ | 269/5000 [04:02<1:16:03, 1.04it/s, beta=1.250, gamma=4.350, train_loss=856.9115, val_loss=892.8193]
Training UniVI: 5%|▌ | 269/5000 [04:03<1:16:03, 1.04it/s, beta=1.250, gamma=4.350, train_loss=855.1410, val_loss=893.7174]
Training UniVI: 5%|▌ | 270/5000 [04:03<1:13:21, 1.07it/s, beta=1.250, gamma=4.350, train_loss=855.1410, val_loss=893.7174]
Training UniVI: 5%|▌ | 270/5000 [04:03<1:13:21, 1.07it/s, beta=1.250, gamma=4.350, train_loss=856.5145, val_loss=893.9444]
Training UniVI: 5%|▌ | 271/5000 [04:03<1:12:02, 1.09it/s, beta=1.250, gamma=4.350, train_loss=856.5145, val_loss=893.9444]
Training UniVI: 5%|▌ | 271/5000 [04:04<1:12:02, 1.09it/s, beta=1.250, gamma=4.350, train_loss=857.3122, val_loss=893.9660]
Training UniVI: 5%|▌ | 272/5000 [04:04<1:14:26, 1.06it/s, beta=1.250, gamma=4.350, train_loss=857.3122, val_loss=893.9660]
Training UniVI: 5%|▌ | 272/5000 [04:05<1:14:26, 1.06it/s, beta=1.250, gamma=4.350, train_loss=856.5295, val_loss=893.1340]
Training UniVI: 5%|▌ | 273/5000 [04:05<1:14:11, 1.06it/s, beta=1.250, gamma=4.350, train_loss=856.5295, val_loss=893.1340]
Training UniVI: 5%|▌ | 273/5000 [04:07<1:14:11, 1.06it/s, beta=1.250, gamma=4.350, train_loss=857.4771, val_loss=892.2536]
Training UniVI: 5%|▌ | 274/5000 [04:07<1:23:38, 1.06s/it, beta=1.250, gamma=4.350, train_loss=857.4771, val_loss=892.2536]
Training UniVI: 5%|▌ | 274/5000 [04:08<1:23:38, 1.06s/it, beta=1.250, gamma=4.350, train_loss=855.2674, val_loss=892.4766]
Training UniVI: 6%|▌ | 275/5000 [04:08<1:21:14, 1.03s/it, beta=1.250, gamma=4.350, train_loss=855.2674, val_loss=892.4766]
Training UniVI: 6%|▌ | 275/5000 [04:09<1:21:14, 1.03s/it, beta=1.250, gamma=4.350, train_loss=855.1741, val_loss=892.6867]
Training UniVI: 6%|▌ | 276/5000 [04:09<1:20:32, 1.02s/it, beta=1.250, gamma=4.350, train_loss=855.1741, val_loss=892.6867]
Training UniVI: 6%|▌ | 276/5000 [04:10<1:20:32, 1.02s/it, beta=1.250, gamma=4.350, train_loss=855.9681, val_loss=893.8149]
Training UniVI: 6%|▌ | 277/5000 [04:10<1:18:47, 1.00s/it, beta=1.250, gamma=4.350, train_loss=855.9681, val_loss=893.8149]
Training UniVI: 6%|▌ | 277/5000 [04:11<1:18:47, 1.00s/it, beta=1.250, gamma=4.350, train_loss=855.1832, val_loss=894.6841]
Training UniVI: 6%|▌ | 278/5000 [04:11<1:17:56, 1.01it/s, beta=1.250, gamma=4.350, train_loss=855.1832, val_loss=894.6841]
Training UniVI: 6%|▌ | 278/5000 [04:12<1:17:56, 1.01it/s, beta=1.250, gamma=4.350, train_loss=854.3954, val_loss=893.1861]
Training UniVI: 6%|▌ | 279/5000 [04:12<1:15:41, 1.04it/s, beta=1.250, gamma=4.350, train_loss=854.3954, val_loss=893.1861]
Training UniVI: 6%|▌ | 279/5000 [04:12<1:15:41, 1.04it/s, beta=1.250, gamma=4.350, train_loss=855.9393, val_loss=893.3358]
Training UniVI: 6%|▌ | 280/5000 [04:12<1:14:00, 1.06it/s, beta=1.250, gamma=4.350, train_loss=855.9393, val_loss=893.3358]
Training UniVI: 6%|▌ | 280/5000 [04:13<1:14:00, 1.06it/s, beta=1.250, gamma=4.350, train_loss=855.8792, val_loss=891.7964]
Training UniVI: 6%|▌ | 281/5000 [04:13<1:12:59, 1.08it/s, beta=1.250, gamma=4.350, train_loss=855.8792, val_loss=891.7964]
Training UniVI: 6%|▌ | 281/5000 [04:14<1:12:59, 1.08it/s, beta=1.250, gamma=4.350, train_loss=855.5994, val_loss=893.0287]
Training UniVI: 6%|▌ | 282/5000 [04:14<1:11:17, 1.10it/s, beta=1.250, gamma=4.350, train_loss=855.5994, val_loss=893.0287]
Training UniVI: 6%|▌ | 282/5000 [04:15<1:11:17, 1.10it/s, beta=1.250, gamma=4.350, train_loss=856.4336, val_loss=891.6010]
Training UniVI: 6%|▌ | 283/5000 [04:15<1:11:32, 1.10it/s, beta=1.250, gamma=4.350, train_loss=856.4336, val_loss=891.6010]
Training UniVI: 6%|▌ | 283/5000 [04:16<1:11:32, 1.10it/s, beta=1.250, gamma=4.350, train_loss=854.9760, val_loss=893.4614]
Training UniVI: 6%|▌ | 284/5000 [04:16<1:09:59, 1.12it/s, beta=1.250, gamma=4.350, train_loss=854.9760, val_loss=893.4614]
Training UniVI: 6%|▌ | 284/5000 [04:17<1:09:59, 1.12it/s, beta=1.250, gamma=4.350, train_loss=853.8397, val_loss=898.8766]
Training UniVI: 6%|▌ | 285/5000 [04:17<1:09:55, 1.12it/s, beta=1.250, gamma=4.350, train_loss=853.8397, val_loss=898.8766]
Training UniVI: 6%|▌ | 285/5000 [04:18<1:09:55, 1.12it/s, beta=1.250, gamma=4.350, train_loss=854.5672, val_loss=892.6690]
Training UniVI: 6%|▌ | 286/5000 [04:18<1:09:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=854.5672, val_loss=892.6690]
Training UniVI: 6%|▌ | 286/5000 [04:19<1:09:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=854.4464, val_loss=894.7972]
Training UniVI: 6%|▌ | 287/5000 [04:19<1:09:01, 1.14it/s, beta=1.250, gamma=4.350, train_loss=854.4464, val_loss=894.7972]
Training UniVI: 6%|▌ | 287/5000 [04:19<1:09:01, 1.14it/s, beta=1.250, gamma=4.350, train_loss=855.2090, val_loss=893.7748]
Training UniVI: 6%|▌ | 288/5000 [04:19<1:09:17, 1.13it/s, beta=1.250, gamma=4.350, train_loss=855.2090, val_loss=893.7748]
Training UniVI: 6%|▌ | 288/5000 [04:20<1:09:17, 1.13it/s, beta=1.250, gamma=4.350, train_loss=854.3831, val_loss=893.4111]
Training UniVI: 6%|▌ | 289/5000 [04:20<1:08:41, 1.14it/s, beta=1.250, gamma=4.350, train_loss=854.3831, val_loss=893.4111]
Training UniVI: 6%|▌ | 289/5000 [04:21<1:08:41, 1.14it/s, beta=1.250, gamma=4.350, train_loss=854.6096, val_loss=895.9566]
Training UniVI: 6%|▌ | 290/5000 [04:21<1:09:24, 1.13it/s, beta=1.250, gamma=4.350, train_loss=854.6096, val_loss=895.9566]
Training UniVI: 6%|▌ | 290/5000 [04:22<1:09:24, 1.13it/s, beta=1.250, gamma=4.350, train_loss=853.6746, val_loss=894.2467]
Training UniVI: 6%|▌ | 291/5000 [04:22<1:08:40, 1.14it/s, beta=1.250, gamma=4.350, train_loss=853.6746, val_loss=894.2467]
Training UniVI: 6%|▌ | 291/5000 [04:23<1:08:40, 1.14it/s, beta=1.250, gamma=4.350, train_loss=855.1457, val_loss=893.4999]
Training UniVI: 6%|▌ | 292/5000 [04:23<1:08:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=855.1457, val_loss=893.4999]
Training UniVI: 6%|▌ | 292/5000 [04:24<1:08:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=852.6382, val_loss=895.0203]
Training UniVI: 6%|▌ | 293/5000 [04:24<1:07:43, 1.16it/s, beta=1.250, gamma=4.350, train_loss=852.6382, val_loss=895.0203]
Training UniVI: 6%|▌ | 293/5000 [04:25<1:07:43, 1.16it/s, beta=1.250, gamma=4.350, train_loss=852.8791, val_loss=894.0383]
Training UniVI: 6%|▌ | 294/5000 [04:25<1:10:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.8791, val_loss=894.0383]
Training UniVI: 6%|▌ | 294/5000 [04:26<1:10:08, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.7441, val_loss=893.5163]
Training UniVI: 6%|▌ | 295/5000 [04:26<1:09:53, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.7441, val_loss=893.5163]
Training UniVI: 6%|▌ | 295/5000 [04:26<1:09:53, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.1140, val_loss=894.9667]
Training UniVI: 6%|▌ | 296/5000 [04:26<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=852.1140, val_loss=894.9667]
Training UniVI: 6%|▌ | 296/5000 [04:27<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=853.0065, val_loss=894.3644]
Training UniVI: 6%|▌ | 297/5000 [04:27<1:09:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=853.0065, val_loss=894.3644]
Training UniVI: 6%|▌ | 297/5000 [04:28<1:09:11, 1.13it/s, beta=1.250, gamma=4.350, train_loss=853.5581, val_loss=895.3234]
Training UniVI: 6%|▌ | 298/5000 [04:28<1:08:46, 1.14it/s, beta=1.250, gamma=4.350, train_loss=853.5581, val_loss=895.3234]
Training UniVI: 6%|▌ | 298/5000 [04:29<1:08:46, 1.14it/s, beta=1.250, gamma=4.350, train_loss=852.4511, val_loss=891.6224]
Training UniVI: 6%|▌ | 299/5000 [04:29<1:08:49, 1.14it/s, beta=1.250, gamma=4.350, train_loss=852.4511, val_loss=891.6224]
[2026-09-21 14:56:21,972] [UniVITrainer] [INFO] [Epoch 300] Train loss=853.5736 (beta=1.250, gamma=4.350)
[2026-09-21 14:56:21,994] [UniVITrainer] [INFO] [Epoch 300] Val loss=892.8429 (beta=1.250, gamma=4.350)
Training UniVI: 6%|▌ | 299/5000 [04:30<1:08:49, 1.14it/s, beta=1.250, gamma=4.350, train_loss=853.5736, val_loss=892.8429]
Training UniVI: 6%|▌ | 300/5000 [04:30<1:09:40, 1.12it/s, beta=1.250, gamma=4.350, train_loss=853.5736, val_loss=892.8429]
Training UniVI: 6%|▌ | 300/5000 [04:31<1:09:40, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.1678, val_loss=893.3083]
Training UniVI: 6%|▌ | 301/5000 [04:31<1:09:09, 1.13it/s, beta=1.250, gamma=4.350, train_loss=852.1678, val_loss=893.3083]
Training UniVI: 6%|▌ | 301/5000 [04:32<1:09:09, 1.13it/s, beta=1.250, gamma=4.350, train_loss=851.8168, val_loss=895.9203]
Training UniVI: 6%|▌ | 302/5000 [04:32<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=851.8168, val_loss=895.9203]
Training UniVI: 6%|▌ | 302/5000 [04:33<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=851.4843, val_loss=894.0059]
Training UniVI: 6%|▌ | 303/5000 [04:33<1:08:03, 1.15it/s, beta=1.250, gamma=4.350, train_loss=851.4843, val_loss=894.0059]
Training UniVI: 6%|▌ | 303/5000 [04:33<1:08:03, 1.15it/s, beta=1.250, gamma=4.350, train_loss=852.8054, val_loss=891.9196]
Training UniVI: 6%|▌ | 304/5000 [04:33<1:08:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=852.8054, val_loss=891.9196]
Training UniVI: 6%|▌ | 304/5000 [04:34<1:08:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=851.8437, val_loss=893.9548]
Training UniVI: 6%|▌ | 305/5000 [04:34<1:07:38, 1.16it/s, beta=1.250, gamma=4.350, train_loss=851.8437, val_loss=893.9548]
Training UniVI: 6%|▌ | 305/5000 [04:35<1:07:38, 1.16it/s, beta=1.250, gamma=4.350, train_loss=852.4247, val_loss=895.5052]
Training UniVI: 6%|▌ | 306/5000 [04:35<1:08:17, 1.15it/s, beta=1.250, gamma=4.350, train_loss=852.4247, val_loss=895.5052]
Training UniVI: 6%|▌ | 306/5000 [04:36<1:08:17, 1.15it/s, beta=1.250, gamma=4.350, train_loss=851.2946, val_loss=895.3180]
Training UniVI: 6%|▌ | 307/5000 [04:36<1:11:18, 1.10it/s, beta=1.250, gamma=4.350, train_loss=851.2946, val_loss=895.3180]
Training UniVI: 6%|▌ | 307/5000 [04:37<1:11:18, 1.10it/s, beta=1.250, gamma=4.350, train_loss=852.7267, val_loss=897.8267]
Training UniVI: 6%|▌ | 308/5000 [04:37<1:09:33, 1.12it/s, beta=1.250, gamma=4.350, train_loss=852.7267, val_loss=897.8267]
Training UniVI: 6%|▌ | 308/5000 [04:38<1:09:33, 1.12it/s, beta=1.250, gamma=4.350, train_loss=851.5317, val_loss=894.8972]
Training UniVI: 6%|▌ | 309/5000 [04:38<1:09:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=851.5317, val_loss=894.8972]
Training UniVI: 6%|▌ | 309/5000 [04:39<1:09:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=852.2939, val_loss=893.8225]
Training UniVI: 6%|▌ | 310/5000 [04:39<1:08:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=852.2939, val_loss=893.8225]
Training UniVI: 6%|▌ | 310/5000 [04:40<1:08:22, 1.14it/s, beta=1.250, gamma=4.350, train_loss=851.8750, val_loss=896.1592]
Training UniVI: 6%|▌ | 311/5000 [04:40<1:08:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=851.8750, val_loss=896.1592]
Training UniVI: 6%|▌ | 311/5000 [04:41<1:08:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=852.0185, val_loss=894.0987]
Training UniVI: 6%|▌ | 312/5000 [04:41<1:08:20, 1.14it/s, beta=1.250, gamma=4.350, train_loss=852.0185, val_loss=894.0987]
Training UniVI: 6%|▌ | 312/5000 [04:41<1:08:20, 1.14it/s, beta=1.250, gamma=4.350, train_loss=850.9218, val_loss=896.4918]
Training UniVI: 6%|▋ | 313/5000 [04:41<1:08:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=850.9218, val_loss=896.4918]
Training UniVI: 6%|▋ | 313/5000 [04:42<1:08:35, 1.14it/s, beta=1.250, gamma=4.350, train_loss=850.5208, val_loss=894.7650]
Training UniVI: 6%|▋ | 314/5000 [04:42<1:09:22, 1.13it/s, beta=1.250, gamma=4.350, train_loss=850.5208, val_loss=894.7650]
Training UniVI: 6%|▋ | 314/5000 [04:43<1:09:22, 1.13it/s, beta=1.250, gamma=4.350, train_loss=850.6843, val_loss=893.7251]
Training UniVI: 6%|▋ | 315/5000 [04:43<1:11:20, 1.09it/s, beta=1.250, gamma=4.350, train_loss=850.6843, val_loss=893.7251]
Training UniVI: 6%|▋ | 315/5000 [04:44<1:11:20, 1.09it/s, beta=1.250, gamma=4.350, train_loss=852.4088, val_loss=895.4644]
Training UniVI: 6%|▋ | 316/5000 [04:44<1:12:55, 1.07it/s, beta=1.250, gamma=4.350, train_loss=852.4088, val_loss=895.4644]
Training UniVI: 6%|▋ | 316/5000 [04:45<1:12:55, 1.07it/s, beta=1.250, gamma=4.350, train_loss=851.5872, val_loss=893.7247]
Training UniVI: 6%|▋ | 317/5000 [04:45<1:11:58, 1.08it/s, beta=1.250, gamma=4.350, train_loss=851.5872, val_loss=893.7247]
Training UniVI: 6%|▋ | 317/5000 [04:46<1:11:58, 1.08it/s, beta=1.250, gamma=4.350, train_loss=851.4669, val_loss=894.0043]
Training UniVI: 6%|▋ | 318/5000 [04:46<1:10:09, 1.11it/s, beta=1.250, gamma=4.350, train_loss=851.4669, val_loss=894.0043]
Training UniVI: 6%|▋ | 318/5000 [04:47<1:10:09, 1.11it/s, beta=1.250, gamma=4.350, train_loss=850.8411, val_loss=894.6975]
Training UniVI: 6%|▋ | 319/5000 [04:47<1:10:40, 1.10it/s, beta=1.250, gamma=4.350, train_loss=850.8411, val_loss=894.6975]
Training UniVI: 6%|▋ | 319/5000 [04:48<1:10:40, 1.10it/s, beta=1.250, gamma=4.350, train_loss=849.7013, val_loss=893.7101]
Training UniVI: 6%|▋ | 320/5000 [04:48<1:10:15, 1.11it/s, beta=1.250, gamma=4.350, train_loss=849.7013, val_loss=893.7101]
Training UniVI: 6%|▋ | 320/5000 [04:49<1:10:15, 1.11it/s, beta=1.250, gamma=4.350, train_loss=849.5194, val_loss=894.3029]
Training UniVI: 6%|▋ | 321/5000 [04:49<1:09:30, 1.12it/s, beta=1.250, gamma=4.350, train_loss=849.5194, val_loss=894.3029]
Training UniVI: 6%|▋ | 321/5000 [04:50<1:09:30, 1.12it/s, beta=1.250, gamma=4.350, train_loss=850.3640, val_loss=893.5768]
Training UniVI: 6%|▋ | 322/5000 [04:50<1:09:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=850.3640, val_loss=893.5768]
Training UniVI: 6%|▋ | 322/5000 [04:50<1:09:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=850.5646, val_loss=894.3408]
Training UniVI: 6%|▋ | 323/5000 [04:50<1:08:21, 1.14it/s, beta=1.250, gamma=4.350, train_loss=850.5646, val_loss=894.3408]
Training UniVI: 6%|▋ | 323/5000 [04:51<1:08:21, 1.14it/s, beta=1.250, gamma=4.350, train_loss=849.8880, val_loss=896.0536]
Training UniVI: 6%|▋ | 324/5000 [04:51<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=849.8880, val_loss=896.0536]
Training UniVI: 6%|▋ | 324/5000 [04:52<1:09:14, 1.13it/s, beta=1.250, gamma=4.350, train_loss=848.9062, val_loss=893.9443]
Training UniVI: 6%|▋ | 325/5000 [04:52<1:08:29, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.9062, val_loss=893.9443]
Training UniVI: 6%|▋ | 325/5000 [04:53<1:08:29, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.4064, val_loss=895.1604]
Training UniVI: 7%|▋ | 326/5000 [04:53<1:09:26, 1.12it/s, beta=1.250, gamma=4.350, train_loss=848.4064, val_loss=895.1604]
Training UniVI: 7%|▋ | 326/5000 [04:54<1:09:26, 1.12it/s, beta=1.250, gamma=4.350, train_loss=849.4627, val_loss=895.5934]
Training UniVI: 7%|▋ | 327/5000 [04:54<1:10:30, 1.10it/s, beta=1.250, gamma=4.350, train_loss=849.4627, val_loss=895.5934]
Training UniVI: 7%|▋ | 327/5000 [04:55<1:10:30, 1.10it/s, beta=1.250, gamma=4.350, train_loss=847.5305, val_loss=893.2341]
Training UniVI: 7%|▋ | 328/5000 [04:55<1:12:05, 1.08it/s, beta=1.250, gamma=4.350, train_loss=847.5305, val_loss=893.2341]
Training UniVI: 7%|▋ | 328/5000 [04:56<1:12:05, 1.08it/s, beta=1.250, gamma=4.350, train_loss=848.9594, val_loss=896.3874]
Training UniVI: 7%|▋ | 329/5000 [04:56<1:10:57, 1.10it/s, beta=1.250, gamma=4.350, train_loss=848.9594, val_loss=896.3874]
Training UniVI: 7%|▋ | 329/5000 [04:57<1:10:57, 1.10it/s, beta=1.250, gamma=4.350, train_loss=849.2012, val_loss=893.4763]
Training UniVI: 7%|▋ | 330/5000 [04:57<1:09:24, 1.12it/s, beta=1.250, gamma=4.350, train_loss=849.2012, val_loss=893.4763]
Training UniVI: 7%|▋ | 330/5000 [04:58<1:09:24, 1.12it/s, beta=1.250, gamma=4.350, train_loss=848.6960, val_loss=892.5109]
Training UniVI: 7%|▋ | 331/5000 [04:58<1:08:48, 1.13it/s, beta=1.250, gamma=4.350, train_loss=848.6960, val_loss=892.5109]
Training UniVI: 7%|▋ | 331/5000 [04:58<1:08:48, 1.13it/s, beta=1.250, gamma=4.350, train_loss=849.2881, val_loss=894.0605]
Training UniVI: 7%|▋ | 332/5000 [04:58<1:07:40, 1.15it/s, beta=1.250, gamma=4.350, train_loss=849.2881, val_loss=894.0605]
Training UniVI: 7%|▋ | 332/5000 [04:59<1:07:40, 1.15it/s, beta=1.250, gamma=4.350, train_loss=850.4830, val_loss=894.3550]
Training UniVI: 7%|▋ | 333/5000 [04:59<1:09:53, 1.11it/s, beta=1.250, gamma=4.350, train_loss=850.4830, val_loss=894.3550]
Training UniVI: 7%|▋ | 333/5000 [05:00<1:09:53, 1.11it/s, beta=1.250, gamma=4.350, train_loss=849.7280, val_loss=894.0816]
Training UniVI: 7%|▋ | 334/5000 [05:00<1:08:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=849.7280, val_loss=894.0816]
Training UniVI: 7%|▋ | 334/5000 [05:01<1:08:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.1540, val_loss=894.3283]
Training UniVI: 7%|▋ | 335/5000 [05:01<1:07:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=848.1540, val_loss=894.3283]
Training UniVI: 7%|▋ | 335/5000 [05:02<1:07:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=847.9839, val_loss=892.9794]
Training UniVI: 7%|▋ | 336/5000 [05:02<1:08:07, 1.14it/s, beta=1.250, gamma=4.350, train_loss=847.9839, val_loss=892.9794]
Training UniVI: 7%|▋ | 336/5000 [05:03<1:08:07, 1.14it/s, beta=1.250, gamma=4.350, train_loss=850.0062, val_loss=895.0888]
Training UniVI: 7%|▋ | 337/5000 [05:03<1:07:25, 1.15it/s, beta=1.250, gamma=4.350, train_loss=850.0062, val_loss=895.0888]
Training UniVI: 7%|▋ | 337/5000 [05:04<1:07:25, 1.15it/s, beta=1.250, gamma=4.350, train_loss=848.8610, val_loss=893.7029]
Training UniVI: 7%|▋ | 338/5000 [05:04<1:06:45, 1.16it/s, beta=1.250, gamma=4.350, train_loss=848.8610, val_loss=893.7029]
Training UniVI: 7%|▋ | 338/5000 [05:05<1:06:45, 1.16it/s, beta=1.250, gamma=4.350, train_loss=848.8555, val_loss=895.4442]
Training UniVI: 7%|▋ | 339/5000 [05:05<1:07:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=848.8555, val_loss=895.4442]
Training UniVI: 7%|▋ | 339/5000 [05:05<1:07:49, 1.15it/s, beta=1.250, gamma=4.350, train_loss=848.2755, val_loss=897.0892]
Training UniVI: 7%|▋ | 340/5000 [05:05<1:06:53, 1.16it/s, beta=1.250, gamma=4.350, train_loss=848.2755, val_loss=897.0892]
Training UniVI: 7%|▋ | 340/5000 [05:06<1:06:53, 1.16it/s, beta=1.250, gamma=4.350, train_loss=847.8539, val_loss=894.7817]
Training UniVI: 7%|▋ | 341/5000 [05:06<1:09:34, 1.12it/s, beta=1.250, gamma=4.350, train_loss=847.8539, val_loss=894.7817]
Training UniVI: 7%|▋ | 341/5000 [05:07<1:09:34, 1.12it/s, beta=1.250, gamma=4.350, train_loss=846.8955, val_loss=893.5989]
Training UniVI: 7%|▋ | 342/5000 [05:07<1:09:43, 1.11it/s, beta=1.250, gamma=4.350, train_loss=846.8955, val_loss=893.5989]
Training UniVI: 7%|▋ | 342/5000 [05:08<1:09:43, 1.11it/s, beta=1.250, gamma=4.350, train_loss=847.5381, val_loss=893.2861]
Training UniVI: 7%|▋ | 343/5000 [05:08<1:08:45, 1.13it/s, beta=1.250, gamma=4.350, train_loss=847.5381, val_loss=893.2861]
Training UniVI: 7%|▋ | 343/5000 [05:09<1:08:45, 1.13it/s, beta=1.250, gamma=4.350, train_loss=848.3061, val_loss=894.3210]
Training UniVI: 7%|▋ | 344/5000 [05:09<1:09:57, 1.11it/s, beta=1.250, gamma=4.350, train_loss=848.3061, val_loss=894.3210]
Training UniVI: 7%|▋ | 344/5000 [05:10<1:09:57, 1.11it/s, beta=1.250, gamma=4.350, train_loss=848.5131, val_loss=896.8990]
Training UniVI: 7%|▋ | 345/5000 [05:10<1:08:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.5131, val_loss=896.8990]
Training UniVI: 7%|▋ | 345/5000 [05:11<1:08:06, 1.14it/s, beta=1.250, gamma=4.350, train_loss=849.1702, val_loss=898.3143]
Training UniVI: 7%|▋ | 346/5000 [05:11<1:08:16, 1.14it/s, beta=1.250, gamma=4.350, train_loss=849.1702, val_loss=898.3143]
Training UniVI: 7%|▋ | 346/5000 [05:12<1:08:16, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.1382, val_loss=896.6802]
Training UniVI: 7%|▋ | 347/5000 [05:12<1:07:48, 1.14it/s, beta=1.250, gamma=4.350, train_loss=848.1382, val_loss=896.6802]
Training UniVI: 7%|▋ | 347/5000 [05:13<1:07:48, 1.14it/s, beta=1.250, gamma=4.350, train_loss=846.9380, val_loss=893.7549]
Training UniVI: 7%|▋ | 348/5000 [05:13<1:08:24, 1.13it/s, beta=1.250, gamma=4.350, train_loss=846.9380, val_loss=893.7549]
Training UniVI: 7%|▋ | 348/5000 [05:13<1:08:24, 1.13it/s, beta=1.250, gamma=4.350, train_loss=848.0925, val_loss=895.0807]
Training UniVI: 7%|▋ | 349/5000 [05:13<1:09:07, 1.12it/s, beta=1.250, gamma=4.350, train_loss=848.0925, val_loss=895.0807]
Training UniVI: 7%|▋ | 349/5000 [05:14<1:09:07, 1.12it/s, beta=1.250, gamma=4.350, train_loss=847.1900, val_loss=893.0070]
Training UniVI: 7%|▋ | 350/5000 [05:14<1:07:49, 1.14it/s, beta=1.250, gamma=4.350, train_loss=847.1900, val_loss=893.0070]
Training UniVI: 7%|▋ | 350/5000 [05:15<1:07:49, 1.14it/s, beta=1.250, gamma=4.350, train_loss=847.1422, val_loss=895.0471]
Training UniVI: 7%|▋ | 351/5000 [05:15<1:08:15, 1.14it/s, beta=1.250, gamma=4.350, train_loss=847.1422, val_loss=895.0471]
Training UniVI: 7%|▋ | 351/5000 [05:16<1:08:15, 1.14it/s, beta=1.250, gamma=4.350, train_loss=847.6239, val_loss=896.0833]
Training UniVI: 7%|▋ | 352/5000 [05:16<1:07:23, 1.15it/s, beta=1.250, gamma=4.350, train_loss=847.6239, val_loss=896.0833]
Training UniVI: 7%|▋ | 352/5000 [05:17<1:07:23, 1.15it/s, beta=1.250, gamma=4.350, train_loss=847.9682, val_loss=894.8434]
Training UniVI: 7%|▋ | 353/5000 [05:17<1:06:20, 1.17it/s, beta=1.250, gamma=4.350, train_loss=847.9682, val_loss=894.8434]
Training UniVI: 7%|▋ | 353/5000 [05:18<1:06:20, 1.17it/s, beta=1.250, gamma=4.350, train_loss=845.7404, val_loss=895.5061]
Training UniVI: 7%|▋ | 354/5000 [05:18<1:06:02, 1.17it/s, beta=1.250, gamma=4.350, train_loss=845.7404, val_loss=895.5061]
Training UniVI: 7%|▋ | 354/5000 [05:19<1:06:02, 1.17it/s, beta=1.250, gamma=4.350, train_loss=846.2384, val_loss=897.2799]
Training UniVI: 7%|▋ | 355/5000 [05:19<1:06:56, 1.16it/s, beta=1.250, gamma=4.350, train_loss=846.2384, val_loss=897.2799]
Training UniVI: 7%|▋ | 355/5000 [05:19<1:06:56, 1.16it/s, beta=1.250, gamma=4.350, train_loss=845.5601, val_loss=895.9292]
Training UniVI: 7%|▋ | 356/5000 [05:19<1:06:13, 1.17it/s, beta=1.250, gamma=4.350, train_loss=845.5601, val_loss=895.9292]
Training UniVI: 7%|▋ | 356/5000 [05:20<1:06:13, 1.17it/s, beta=1.250, gamma=4.350, train_loss=845.6470, val_loss=895.5358]
Training UniVI: 7%|▋ | 357/5000 [05:20<1:06:07, 1.17it/s, beta=1.250, gamma=4.350, train_loss=845.6470, val_loss=895.5358]
Training UniVI: 7%|▋ | 357/5000 [05:21<1:06:07, 1.17it/s, beta=1.250, gamma=4.350, train_loss=846.9450, val_loss=894.6926]
Training UniVI: 7%|▋ | 358/5000 [05:21<1:07:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=846.9450, val_loss=894.6926]
Training UniVI: 7%|▋ | 358/5000 [05:22<1:07:10, 1.15it/s, beta=1.250, gamma=4.350, train_loss=848.3147, val_loss=897.5345]
Training UniVI: 7%|▋ | 359/5000 [05:22<1:06:22, 1.17it/s, beta=1.250, gamma=4.350, train_loss=848.3147, val_loss=897.5345]
Training UniVI: 7%|▋ | 359/5000 [05:23<1:06:22, 1.17it/s, beta=1.250, gamma=4.350, train_loss=846.1181, val_loss=893.1842]
Training UniVI: 7%|▋ | 360/5000 [05:23<1:06:14, 1.17it/s, beta=1.250, gamma=4.350, train_loss=846.1181, val_loss=893.1842]
Training UniVI: 7%|▋ | 360/5000 [05:24<1:06:14, 1.17it/s, beta=1.250, gamma=4.350, train_loss=846.5589, val_loss=895.6470]
Training UniVI: 7%|▋ | 361/5000 [05:24<1:05:43, 1.18it/s, beta=1.250, gamma=4.350, train_loss=846.5589, val_loss=895.6470]
Training UniVI: 7%|▋ | 361/5000 [05:25<1:05:43, 1.18it/s, beta=1.250, gamma=4.350, train_loss=847.2730, val_loss=895.2081]
Training UniVI: 7%|▋ | 362/5000 [05:25<1:08:36, 1.13it/s, beta=1.250, gamma=4.350, train_loss=847.2730, val_loss=895.2081]
Training UniVI: 7%|▋ | 362/5000 [05:26<1:08:36, 1.13it/s, beta=1.250, gamma=4.350, train_loss=847.9959, val_loss=893.3555]
Training UniVI: 7%|▋ | 363/5000 [05:26<1:08:28, 1.13it/s, beta=1.250, gamma=4.350, train_loss=847.9959, val_loss=893.3555]
Training UniVI: 7%|▋ | 363/5000 [05:27<1:08:28, 1.13it/s, beta=1.250, gamma=4.350, train_loss=846.2641, val_loss=893.7079]
Training UniVI: 7%|▋ | 364/5000 [05:27<1:10:04, 1.10it/s, beta=1.250, gamma=4.350, train_loss=846.2641, val_loss=893.7079]
Training UniVI: 7%|▋ | 364/5000 [05:27<1:10:04, 1.10it/s, beta=1.250, gamma=4.350, train_loss=846.2335, val_loss=897.1877]
Training UniVI: 7%|▋ | 365/5000 [05:27<1:08:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=846.2335, val_loss=897.1877]
Training UniVI: 7%|▋ | 365/5000 [05:28<1:08:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=846.0341, val_loss=895.4381]
Training UniVI: 7%|▋ | 366/5000 [05:28<1:09:34, 1.11it/s, beta=1.250, gamma=4.350, train_loss=846.0341, val_loss=895.4381]
Training UniVI: 7%|▋ | 366/5000 [05:29<1:09:34, 1.11it/s, beta=1.250, gamma=4.350, train_loss=846.2812, val_loss=894.3188]
Training UniVI: 7%|▋ | 367/5000 [05:29<1:08:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=846.2812, val_loss=894.3188]
Training UniVI: 7%|▋ | 367/5000 [05:30<1:08:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=845.5234, val_loss=895.1610]
Training UniVI: 7%|▋ | 368/5000 [05:30<1:07:19, 1.15it/s, beta=1.250, gamma=4.350, train_loss=845.5234, val_loss=895.1610]
Training UniVI: 7%|▋ | 368/5000 [05:31<1:07:19, 1.15it/s, beta=1.250, gamma=4.350, train_loss=845.0282, val_loss=896.0759]
Training UniVI: 7%|▋ | 369/5000 [05:31<1:06:48, 1.16it/s, beta=1.250, gamma=4.350, train_loss=845.0282, val_loss=896.0759]
Training UniVI: 7%|▋ | 369/5000 [05:32<1:06:48, 1.16it/s, beta=1.250, gamma=4.350, train_loss=845.6563, val_loss=897.5500]
Training UniVI: 7%|▋ | 370/5000 [05:32<1:07:20, 1.15it/s, beta=1.250, gamma=4.350, train_loss=845.6563, val_loss=897.5500]
Training UniVI: 7%|▋ | 370/5000 [05:33<1:07:20, 1.15it/s, beta=1.250, gamma=4.350, train_loss=846.2029, val_loss=895.9276]
Training UniVI: 7%|▋ | 371/5000 [05:33<1:06:52, 1.15it/s, beta=1.250, gamma=4.350, train_loss=846.2029, val_loss=895.9276]
Training UniVI: 7%|▋ | 371/5000 [05:33<1:06:52, 1.15it/s, beta=1.250, gamma=4.350, train_loss=845.5968, val_loss=895.6357]
Training UniVI: 7%|▋ | 372/5000 [05:33<1:06:21, 1.16it/s, beta=1.250, gamma=4.350, train_loss=845.5968, val_loss=895.6357]
Training UniVI: 7%|▋ | 372/5000 [05:34<1:06:21, 1.16it/s, beta=1.250, gamma=4.350, train_loss=845.1743, val_loss=897.3396]
Training UniVI: 7%|▋ | 373/5000 [05:34<1:07:48, 1.14it/s, beta=1.250, gamma=4.350, train_loss=845.1743, val_loss=897.3396]
Training UniVI: 7%|▋ | 373/5000 [05:35<1:07:48, 1.14it/s, beta=1.250, gamma=4.350, train_loss=844.8563, val_loss=895.9424]
Training UniVI: 7%|▋ | 374/5000 [05:35<1:08:06, 1.13it/s, beta=1.250, gamma=4.350, train_loss=844.8563, val_loss=895.9424]
Training UniVI: 7%|▋ | 374/5000 [05:36<1:08:06, 1.13it/s, beta=1.250, gamma=4.350, train_loss=845.7222, val_loss=897.8293]
Training UniVI: 8%|▊ | 375/5000 [05:36<1:11:42, 1.07it/s, beta=1.250, gamma=4.350, train_loss=845.7222, val_loss=897.8293]
Training UniVI: 8%|▊ | 375/5000 [05:37<1:11:42, 1.07it/s, beta=1.250, gamma=4.350, train_loss=845.3154, val_loss=894.9247]
Training UniVI: 8%|▊ | 376/5000 [05:37<1:15:17, 1.02it/s, beta=1.250, gamma=4.350, train_loss=845.3154, val_loss=894.9247]
Training UniVI: 8%|▊ | 376/5000 [05:38<1:15:17, 1.02it/s, beta=1.250, gamma=4.350, train_loss=843.6252, val_loss=894.5220]
Training UniVI: 8%|▊ | 377/5000 [05:38<1:14:32, 1.03it/s, beta=1.250, gamma=4.350, train_loss=843.6252, val_loss=894.5220]
Training UniVI: 8%|▊ | 377/5000 [05:39<1:14:32, 1.03it/s, beta=1.250, gamma=4.350, train_loss=844.3553, val_loss=895.5202]
Training UniVI: 8%|▊ | 378/5000 [05:39<1:12:18, 1.07it/s, beta=1.250, gamma=4.350, train_loss=844.3553, val_loss=895.5202]
Training UniVI: 8%|▊ | 378/5000 [05:40<1:12:18, 1.07it/s, beta=1.250, gamma=4.350, train_loss=845.5491, val_loss=895.0709]
Training UniVI: 8%|▊ | 379/5000 [05:40<1:12:30, 1.06it/s, beta=1.250, gamma=4.350, train_loss=845.5491, val_loss=895.0709]
Training UniVI: 8%|▊ | 379/5000 [05:41<1:12:30, 1.06it/s, beta=1.250, gamma=4.350, train_loss=845.8209, val_loss=899.2132]
Training UniVI: 8%|▊ | 380/5000 [05:41<1:11:25, 1.08it/s, beta=1.250, gamma=4.350, train_loss=845.8209, val_loss=899.2132]
Training UniVI: 8%|▊ | 380/5000 [05:42<1:11:25, 1.08it/s, beta=1.250, gamma=4.350, train_loss=844.4252, val_loss=895.6632]
Training UniVI: 8%|▊ | 381/5000 [05:42<1:11:05, 1.08it/s, beta=1.250, gamma=4.350, train_loss=844.4252, val_loss=895.6632]
Training UniVI: 8%|▊ | 381/5000 [05:43<1:11:05, 1.08it/s, beta=1.250, gamma=4.350, train_loss=842.8750, val_loss=899.9986]
Training UniVI: 8%|▊ | 382/5000 [05:43<1:11:14, 1.08it/s, beta=1.250, gamma=4.350, train_loss=842.8750, val_loss=899.9986]
Training UniVI: 8%|▊ | 382/5000 [05:44<1:11:14, 1.08it/s, beta=1.250, gamma=4.350, train_loss=842.1766, val_loss=895.7039]
Training UniVI: 8%|▊ | 383/5000 [05:44<1:12:01, 1.07it/s, beta=1.250, gamma=4.350, train_loss=842.1766, val_loss=895.7039]
Training UniVI: 8%|▊ | 383/5000 [05:45<1:12:01, 1.07it/s, beta=1.250, gamma=4.350, train_loss=843.2452, val_loss=895.2832]
Training UniVI: 8%|▊ | 384/5000 [05:45<1:11:49, 1.07it/s, beta=1.250, gamma=4.350, train_loss=843.2452, val_loss=895.2832]
Training UniVI: 8%|▊ | 384/5000 [05:46<1:11:49, 1.07it/s, beta=1.250, gamma=4.350, train_loss=843.5701, val_loss=897.2200]
Training UniVI: 8%|▊ | 385/5000 [05:46<1:11:32, 1.08it/s, beta=1.250, gamma=4.350, train_loss=843.5701, val_loss=897.2200]
Training UniVI: 8%|▊ | 385/5000 [05:47<1:11:32, 1.08it/s, beta=1.250, gamma=4.350, train_loss=843.7744, val_loss=895.6052]
Training UniVI: 8%|▊ | 386/5000 [05:47<1:10:46, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.7744, val_loss=895.6052]
Training UniVI: 8%|▊ | 386/5000 [05:48<1:10:46, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.6030, val_loss=896.8842]
Training UniVI: 8%|▊ | 387/5000 [05:48<1:11:49, 1.07it/s, beta=1.250, gamma=4.350, train_loss=844.6030, val_loss=896.8842]
Training UniVI: 8%|▊ | 387/5000 [05:49<1:11:49, 1.07it/s, beta=1.250, gamma=4.350, train_loss=844.2127, val_loss=896.2704]
Training UniVI: 8%|▊ | 388/5000 [05:49<1:12:08, 1.07it/s, beta=1.250, gamma=4.350, train_loss=844.2127, val_loss=896.2704]
Training UniVI: 8%|▊ | 388/5000 [05:49<1:12:08, 1.07it/s, beta=1.250, gamma=4.350, train_loss=844.3907, val_loss=895.9094]
Training UniVI: 8%|▊ | 389/5000 [05:49<1:10:41, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.3907, val_loss=895.9094]
Training UniVI: 8%|▊ | 389/5000 [05:50<1:10:41, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.2918, val_loss=896.4838]
Training UniVI: 8%|▊ | 390/5000 [05:50<1:10:48, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.2918, val_loss=896.4838]
Training UniVI: 8%|▊ | 390/5000 [05:51<1:10:48, 1.09it/s, beta=1.250, gamma=4.350, train_loss=841.4031, val_loss=896.0655]
Training UniVI: 8%|▊ | 391/5000 [05:51<1:09:40, 1.10it/s, beta=1.250, gamma=4.350, train_loss=841.4031, val_loss=896.0655]
Training UniVI: 8%|▊ | 391/5000 [05:52<1:09:40, 1.10it/s, beta=1.250, gamma=4.350, train_loss=843.7633, val_loss=895.1007]
Training UniVI: 8%|▊ | 392/5000 [05:52<1:08:33, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.7633, val_loss=895.1007]
Training UniVI: 8%|▊ | 392/5000 [05:53<1:08:33, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.3149, val_loss=894.4037]
Training UniVI: 8%|▊ | 393/5000 [05:53<1:10:40, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.3149, val_loss=894.4037]
Training UniVI: 8%|▊ | 393/5000 [05:54<1:10:40, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.2241, val_loss=895.8311]
Training UniVI: 8%|▊ | 394/5000 [05:54<1:10:06, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.2241, val_loss=895.8311]
Training UniVI: 8%|▊ | 394/5000 [05:55<1:10:06, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.1166, val_loss=894.9290]
Training UniVI: 8%|▊ | 395/5000 [05:55<1:10:12, 1.09it/s, beta=1.250, gamma=4.350, train_loss=844.1166, val_loss=894.9290]
Training UniVI: 8%|▊ | 395/5000 [05:56<1:10:12, 1.09it/s, beta=1.250, gamma=4.350, train_loss=842.3233, val_loss=894.3187]
Training UniVI: 8%|▊ | 396/5000 [05:56<1:10:42, 1.09it/s, beta=1.250, gamma=4.350, train_loss=842.3233, val_loss=894.3187]
Training UniVI: 8%|▊ | 396/5000 [05:57<1:10:42, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.4303, val_loss=895.7432]
Training UniVI: 8%|▊ | 397/5000 [05:57<1:09:15, 1.11it/s, beta=1.250, gamma=4.350, train_loss=843.4303, val_loss=895.7432]
Training UniVI: 8%|▊ | 397/5000 [05:58<1:09:15, 1.11it/s, beta=1.250, gamma=4.350, train_loss=842.2548, val_loss=897.9800]
Training UniVI: 8%|▊ | 398/5000 [05:58<1:08:27, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.2548, val_loss=897.9800]
Training UniVI: 8%|▊ | 398/5000 [05:58<1:08:27, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.9361, val_loss=896.8043]
Training UniVI: 8%|▊ | 399/5000 [05:58<1:08:42, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.9361, val_loss=896.8043]
[2026-09-21 14:57:51,227] [UniVITrainer] [INFO] [Epoch 400] Train loss=840.5433 (beta=1.250, gamma=4.350)
[2026-09-21 14:57:51,250] [UniVITrainer] [INFO] [Epoch 400] Val loss=894.4801 (beta=1.250, gamma=4.350)
Training UniVI: 8%|▊ | 399/5000 [05:59<1:08:42, 1.12it/s, beta=1.250, gamma=4.350, train_loss=840.5433, val_loss=894.4801]
Training UniVI: 8%|▊ | 400/5000 [05:59<1:07:30, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.5433, val_loss=894.4801]
Training UniVI: 8%|▊ | 400/5000 [06:00<1:07:30, 1.14it/s, beta=1.250, gamma=4.350, train_loss=842.0710, val_loss=894.9406]
Training UniVI: 8%|▊ | 401/5000 [06:00<1:06:36, 1.15it/s, beta=1.250, gamma=4.350, train_loss=842.0710, val_loss=894.9406]
Training UniVI: 8%|▊ | 401/5000 [06:01<1:06:36, 1.15it/s, beta=1.250, gamma=4.350, train_loss=843.9783, val_loss=897.1451]
Training UniVI: 8%|▊ | 402/5000 [06:01<1:08:15, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.9783, val_loss=897.1451]
Training UniVI: 8%|▊ | 402/5000 [06:02<1:08:15, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.8836, val_loss=897.1552]
Training UniVI: 8%|▊ | 403/5000 [06:02<1:08:06, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.8836, val_loss=897.1552]
Training UniVI: 8%|▊ | 403/5000 [06:03<1:08:06, 1.12it/s, beta=1.250, gamma=4.350, train_loss=840.8686, val_loss=893.7039]
Training UniVI: 8%|▊ | 404/5000 [06:03<1:07:42, 1.13it/s, beta=1.250, gamma=4.350, train_loss=840.8686, val_loss=893.7039]
Training UniVI: 8%|▊ | 404/5000 [06:04<1:07:42, 1.13it/s, beta=1.250, gamma=4.350, train_loss=844.1363, val_loss=893.8013]
Training UniVI: 8%|▊ | 405/5000 [06:04<1:07:48, 1.13it/s, beta=1.250, gamma=4.350, train_loss=844.1363, val_loss=893.8013]
Training UniVI: 8%|▊ | 405/5000 [06:05<1:07:48, 1.13it/s, beta=1.250, gamma=4.350, train_loss=842.7289, val_loss=894.6885]
Training UniVI: 8%|▊ | 406/5000 [06:05<1:07:27, 1.13it/s, beta=1.250, gamma=4.350, train_loss=842.7289, val_loss=894.6885]
Training UniVI: 8%|▊ | 406/5000 [06:05<1:07:27, 1.13it/s, beta=1.250, gamma=4.350, train_loss=843.6339, val_loss=896.8179]
Training UniVI: 8%|▊ | 407/5000 [06:05<1:08:25, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.6339, val_loss=896.8179]
Training UniVI: 8%|▊ | 407/5000 [06:06<1:08:25, 1.12it/s, beta=1.250, gamma=4.350, train_loss=843.0467, val_loss=895.6278]
Training UniVI: 8%|▊ | 408/5000 [06:06<1:10:28, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.0467, val_loss=895.6278]
Training UniVI: 8%|▊ | 408/5000 [06:07<1:10:28, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.1231, val_loss=893.5955]
Training UniVI: 8%|▊ | 409/5000 [06:07<1:09:59, 1.09it/s, beta=1.250, gamma=4.350, train_loss=843.1231, val_loss=893.5955]
Training UniVI: 8%|▊ | 409/5000 [06:08<1:09:59, 1.09it/s, beta=1.250, gamma=4.350, train_loss=841.4232, val_loss=896.3432]
Training UniVI: 8%|▊ | 410/5000 [06:08<1:09:10, 1.11it/s, beta=1.250, gamma=4.350, train_loss=841.4232, val_loss=896.3432]
Training UniVI: 8%|▊ | 410/5000 [06:09<1:09:10, 1.11it/s, beta=1.250, gamma=4.350, train_loss=840.9792, val_loss=897.5577]
Training UniVI: 8%|▊ | 411/5000 [06:09<1:08:11, 1.12it/s, beta=1.250, gamma=4.350, train_loss=840.9792, val_loss=897.5577]
Training UniVI: 8%|▊ | 411/5000 [06:10<1:08:11, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.3035, val_loss=897.8556]
Training UniVI: 8%|▊ | 412/5000 [06:10<1:08:10, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.3035, val_loss=897.8556]
Training UniVI: 8%|▊ | 412/5000 [06:11<1:08:10, 1.12it/s, beta=1.250, gamma=4.350, train_loss=841.2943, val_loss=897.6082]
Training UniVI: 8%|▊ | 413/5000 [06:11<1:09:16, 1.10it/s, beta=1.250, gamma=4.350, train_loss=841.2943, val_loss=897.6082]
Training UniVI: 8%|▊ | 413/5000 [06:12<1:09:16, 1.10it/s, beta=1.250, gamma=4.350, train_loss=840.3856, val_loss=896.4655]
Training UniVI: 8%|▊ | 414/5000 [06:12<1:08:32, 1.12it/s, beta=1.250, gamma=4.350, train_loss=840.3856, val_loss=896.4655]
Training UniVI: 8%|▊ | 414/5000 [06:13<1:08:32, 1.12it/s, beta=1.250, gamma=4.350, train_loss=840.4323, val_loss=895.7471]
Training UniVI: 8%|▊ | 415/5000 [06:13<1:10:07, 1.09it/s, beta=1.250, gamma=4.350, train_loss=840.4323, val_loss=895.7471]
Training UniVI: 8%|▊ | 415/5000 [06:14<1:10:07, 1.09it/s, beta=1.250, gamma=4.350, train_loss=841.0766, val_loss=897.6195]
Training UniVI: 8%|▊ | 416/5000 [06:14<1:08:41, 1.11it/s, beta=1.250, gamma=4.350, train_loss=841.0766, val_loss=897.6195]
Training UniVI: 8%|▊ | 416/5000 [06:14<1:08:41, 1.11it/s, beta=1.250, gamma=4.350, train_loss=840.2462, val_loss=895.7235]
Training UniVI: 8%|▊ | 417/5000 [06:14<1:07:28, 1.13it/s, beta=1.250, gamma=4.350, train_loss=840.2462, val_loss=895.7235]
Training UniVI: 8%|▊ | 417/5000 [06:15<1:07:28, 1.13it/s, beta=1.250, gamma=4.350, train_loss=841.3321, val_loss=899.1304]
Training UniVI: 8%|▊ | 418/5000 [06:15<1:07:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=841.3321, val_loss=899.1304]
Training UniVI: 8%|▊ | 418/5000 [06:16<1:07:34, 1.13it/s, beta=1.250, gamma=4.350, train_loss=842.2327, val_loss=898.6735]
Training UniVI: 8%|▊ | 419/5000 [06:16<1:06:40, 1.15it/s, beta=1.250, gamma=4.350, train_loss=842.2327, val_loss=898.6735]
Training UniVI: 8%|▊ | 419/5000 [06:17<1:06:40, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.4745, val_loss=899.8841]
Training UniVI: 8%|▊ | 420/5000 [06:17<1:07:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=840.4745, val_loss=899.8841]
Training UniVI: 8%|▊ | 420/5000 [06:18<1:07:30, 1.13it/s, beta=1.250, gamma=4.350, train_loss=841.2573, val_loss=896.2463]
Training UniVI: 8%|▊ | 421/5000 [06:18<1:06:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=841.2573, val_loss=896.2463]
Training UniVI: 8%|▊ | 421/5000 [06:19<1:06:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=841.0099, val_loss=896.3029]
Training UniVI: 8%|▊ | 422/5000 [06:19<1:06:47, 1.14it/s, beta=1.250, gamma=4.350, train_loss=841.0099, val_loss=896.3029]
Training UniVI: 8%|▊ | 422/5000 [06:20<1:06:47, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.8267, val_loss=898.7378]
Training UniVI: 8%|▊ | 423/5000 [06:20<1:06:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.8267, val_loss=898.7378]
Training UniVI: 8%|▊ | 423/5000 [06:21<1:06:28, 1.15it/s, beta=1.250, gamma=4.350, train_loss=839.7775, val_loss=899.5505]
Training UniVI: 8%|▊ | 424/5000 [06:21<1:06:42, 1.14it/s, beta=1.250, gamma=4.350, train_loss=839.7775, val_loss=899.5505]
Training UniVI: 8%|▊ | 424/5000 [06:21<1:06:42, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.3182, val_loss=898.2443]
Training UniVI: 8%|▊ | 425/5000 [06:21<1:05:46, 1.16it/s, beta=1.250, gamma=4.350, train_loss=840.3182, val_loss=898.2443]
Training UniVI: 8%|▊ | 425/5000 [06:22<1:05:46, 1.16it/s, beta=1.250, gamma=4.350, train_loss=840.4197, val_loss=898.7586]
Training UniVI: 9%|▊ | 426/5000 [06:22<1:05:32, 1.16it/s, beta=1.250, gamma=4.350, train_loss=840.4197, val_loss=898.7586]
Training UniVI: 9%|▊ | 426/5000 [06:23<1:05:32, 1.16it/s, beta=1.250, gamma=4.350, train_loss=840.0056, val_loss=898.9640]
Training UniVI: 9%|▊ | 427/5000 [06:23<1:06:40, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.0056, val_loss=898.9640]
Training UniVI: 9%|▊ | 427/5000 [06:24<1:06:40, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.7039, val_loss=894.9705]
Training UniVI: 9%|▊ | 428/5000 [06:24<1:06:05, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.7039, val_loss=894.9705]
Training UniVI: 9%|▊ | 428/5000 [06:25<1:06:05, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.3323, val_loss=895.4365]
Training UniVI: 9%|▊ | 429/5000 [06:25<1:07:17, 1.13it/s, beta=1.250, gamma=4.350, train_loss=840.3323, val_loss=895.4365]
Training UniVI: 9%|▊ | 429/5000 [06:26<1:07:17, 1.13it/s, beta=1.250, gamma=4.350, train_loss=842.9233, val_loss=899.2620]
Training UniVI: 9%|▊ | 430/5000 [06:26<1:06:30, 1.15it/s, beta=1.250, gamma=4.350, train_loss=842.9233, val_loss=899.2620]
Training UniVI: 9%|▊ | 430/5000 [06:27<1:06:30, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.7436, val_loss=897.1481]
Training UniVI: 9%|▊ | 431/5000 [06:27<1:06:27, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.7436, val_loss=897.1481]
Training UniVI: 9%|▊ | 431/5000 [06:28<1:06:27, 1.15it/s, beta=1.250, gamma=4.350, train_loss=842.2519, val_loss=896.7901]
Training UniVI: 9%|▊ | 432/5000 [06:28<1:08:16, 1.12it/s, beta=1.250, gamma=4.350, train_loss=842.2519, val_loss=896.7901]
Training UniVI: 9%|▊ | 432/5000 [06:29<1:08:16, 1.12it/s, beta=1.250, gamma=4.350, train_loss=837.9045, val_loss=896.6564]
Training UniVI: 9%|▊ | 433/5000 [06:29<1:09:18, 1.10it/s, beta=1.250, gamma=4.350, train_loss=837.9045, val_loss=896.6564]
Training UniVI: 9%|▊ | 433/5000 [06:29<1:09:18, 1.10it/s, beta=1.250, gamma=4.350, train_loss=841.7964, val_loss=896.8764]
Training UniVI: 9%|▊ | 434/5000 [06:29<1:08:56, 1.10it/s, beta=1.250, gamma=4.350, train_loss=841.7964, val_loss=896.8764]
Training UniVI: 9%|▊ | 434/5000 [06:30<1:08:56, 1.10it/s, beta=1.250, gamma=4.350, train_loss=840.5582, val_loss=895.6222]
Training UniVI: 9%|▊ | 435/5000 [06:30<1:08:16, 1.11it/s, beta=1.250, gamma=4.350, train_loss=840.5582, val_loss=895.6222]
Training UniVI: 9%|▊ | 435/5000 [06:31<1:08:16, 1.11it/s, beta=1.250, gamma=4.350, train_loss=839.7584, val_loss=896.3117]
Training UniVI: 9%|▊ | 436/5000 [06:31<1:07:21, 1.13it/s, beta=1.250, gamma=4.350, train_loss=839.7584, val_loss=896.3117]
Training UniVI: 9%|▊ | 436/5000 [06:32<1:07:21, 1.13it/s, beta=1.250, gamma=4.350, train_loss=839.3541, val_loss=896.2592]
Training UniVI: 9%|▊ | 437/5000 [06:32<1:06:18, 1.15it/s, beta=1.250, gamma=4.350, train_loss=839.3541, val_loss=896.2592]
Training UniVI: 9%|▊ | 437/5000 [06:33<1:06:18, 1.15it/s, beta=1.250, gamma=4.350, train_loss=839.6951, val_loss=896.2981]
Training UniVI: 9%|▉ | 438/5000 [06:33<1:07:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=839.6951, val_loss=896.2981]
Training UniVI: 9%|▉ | 438/5000 [06:34<1:07:16, 1.13it/s, beta=1.250, gamma=4.350, train_loss=840.5742, val_loss=895.9615]
Training UniVI: 9%|▉ | 439/5000 [06:34<1:06:33, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.5742, val_loss=895.9615]
Training UniVI: 9%|▉ | 439/5000 [06:35<1:06:33, 1.14it/s, beta=1.250, gamma=4.350, train_loss=840.7601, val_loss=896.5856]
Training UniVI: 9%|▉ | 440/5000 [06:35<1:06:18, 1.15it/s, beta=1.250, gamma=4.350, train_loss=840.7601, val_loss=896.5856]
Training UniVI: 9%|▉ | 440/5000 [06:36<1:06:18, 1.15it/s, beta=1.250, gamma=4.350, train_loss=838.9070, val_loss=895.0868]
Training UniVI: 9%|▉ | 441/5000 [06:36<1:07:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=838.9070, val_loss=895.0868]
Training UniVI: 9%|▉ | 441/5000 [06:37<1:07:20, 1.13it/s, beta=1.250, gamma=4.350, train_loss=839.6657, val_loss=892.5396]
Training UniVI: 9%|▉ | 442/5000 [06:37<1:08:46, 1.10it/s, beta=1.250, gamma=4.350, train_loss=839.6657, val_loss=892.5396]
Training UniVI: 9%|▉ | 442/5000 [06:37<1:08:46, 1.10it/s, beta=1.250, gamma=4.350, train_loss=839.5403, val_loss=895.4877]
Training UniVI: 9%|▉ | 443/5000 [06:37<1:08:24, 1.11it/s, beta=1.250, gamma=4.350, train_loss=839.5403, val_loss=895.4877]
Training UniVI: 9%|▉ | 443/5000 [06:38<1:08:24, 1.11it/s, beta=1.250, gamma=4.350, train_loss=839.4897, val_loss=898.6407]
Training UniVI: 9%|▉ | 444/5000 [06:38<1:07:45, 1.12it/s, beta=1.250, gamma=4.350, train_loss=839.4897, val_loss=898.6407]
Training UniVI: 9%|▉ | 444/5000 [06:39<1:07:45, 1.12it/s, beta=1.250, gamma=4.350, train_loss=838.7522, val_loss=896.0678]
Training UniVI: 9%|▉ | 445/5000 [06:39<1:06:45, 1.14it/s, beta=1.250, gamma=4.350, train_loss=838.7522, val_loss=896.0678]
Training UniVI: 9%|▉ | 445/5000 [06:40<1:06:45, 1.14it/s, beta=1.250, gamma=4.350, train_loss=839.3402, val_loss=896.8531]
Training UniVI: 9%|▉ | 446/5000 [06:40<1:05:43, 1.15it/s, beta=1.250, gamma=4.350, train_loss=839.3402, val_loss=896.8531]
Training UniVI: 9%|▉ | 446/5000 [06:41<1:05:43, 1.15it/s, beta=1.250, gamma=4.350, train_loss=838.1211, val_loss=894.8652]
Training UniVI: 9%|▉ | 447/5000 [06:41<1:06:25, 1.14it/s, beta=1.250, gamma=4.350, train_loss=838.1211, val_loss=894.8652]
Training UniVI: 9%|▉ | 447/5000 [06:42<1:06:25, 1.14it/s, beta=1.250, gamma=4.350, train_loss=839.2747, val_loss=896.3339]
[2026-09-21 14:58:33,741] [UniVITrainer] [INFO] Early stopping at epoch 448 (best val loss=886.8277, best epoch=148)
Training UniVI: 9%|▉ | 447/5000 [06:42<1:08:17, 1.11it/s, beta=1.250, gamma=4.350, train_loss=839.2747, val_loss=896.3339]
[2026-09-21 14:58:33,746] [UniVITrainer] [INFO] Restored best model from epoch 148 (val loss=886.8277)
Held-out alignment (Fig. 4)#
test = parts["test"]
z_rna = encode_adata(model, test["rna"], modality="rna", device=device, latent="modality_mean")
z_atac = encode_adata(model, test["atac"], modality="atac", device=device, latent="modality_mean")
lab = test["rna"].obs["cell_type"].astype(str).to_numpy()
m = evaluate_alignment(Z1=z_rna, Z2=z_atac, labels_source=lab, labels_target=lab, recall_ks=(1, 10, 50))
pd.Series({
"FOSCTTM": m["foscttm_mean"],
"Recall@1": m["recall_at_k"]["1"]["mean"], "Recall@10": m["recall_at_k"]["10"]["mean"],
"Recall@50": m["recall_at_k"]["50"]["mean"],
"label transfer acc (RNA to ATAC)": m["label_transfer_acc"],
"macro-F1, worse direction": m["worst_direction_macro_f1"],
"modality mixing (k=20)": compute_modality_mixing(np.vstack([z_rna, z_atac]), np.repeat(["rna", "atac"], len(z_rna)), k=20),
}).round(3)
FOSCTTM 0.048
Recall@1 0.037
Recall@10 0.183
Recall@50 0.425
label transfer acc (RNA to ATAC) 0.963
macro-F1, worse direction 0.722
modality mixing (k=20) 0.192
dtype: float64
joint = stack_embeddings(model, [("test", "rna", test["rna"]), ("test", "atac", test["atac"])], device=device)
sc.pp.neighbors(joint, use_rep="X_univi", n_neighbors=30)
sc.tl.umap(joint, random_state=0)
sc.pl.umap(joint, color=["modality", "cell_type"], wspace=0.45, legend_fontsize=6)
Three-dimensional views of the same embedding (Supplemental Fig. S2)#
joint3 = joint.copy()
sc.tl.umap(joint3, n_components=3, random_state=0)
sc.pl.umap(joint3, color=["modality", "cell_type"], projection="3d", wspace=0.3, legend_fontsize=6)
ATAC to RNA prediction of marker genes#
rna_t = test["rna"]
rna_t.layers["from_atac"] = cross_modal_predict(model, test["atac"], src_mod="atac", tgt_mod="rna", device=device)
rna_t.obsm["X_univi"] = z_rna
sc.pp.neighbors(rna_t, use_rep="X_univi")
sc.tl.umap(rna_t, random_state=0)
genes = [g for g in ["MS4A1", "CD3D", "NKG7", "LYZ", "FCGR3A", "IL7R"] if g in rna_t.var_names]
sc.pl.umap(rna_t, color=genes, ncols=3, vmax="p99", cmap="viridis", title=[f"{g} observed" for g in genes])
sc.pl.umap(rna_t, color=genes, ncols=3, layer="from_atac", vmax="p99", cmap="viridis", title=[f"{g} from ATAC" for g in genes])
