Downloads · 30 days
0
lhx05/lab1-cvlface-code
lab1-cvlface-code is a machine learning model from lhx05. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as unknown.
本仓库用于备份 Lab1 人脸识别实验中使用的 CVLFace 代码和训练脚本。
Downloads · 30 days
0
Access
Public
Updated Apr 16, 2026
Repo size
—
Likes
0
Public
Click a slice to open those files.
.py694 KB · 84%
From the Hugging Face model README
本仓库用于备份 Lab1 人脸识别实验中使用的 CVLFace 代码和训练脚本。
代码来自服务器实验目录:
/root/Lab1/project/cvlface
上传时已经排除数据集、实验输出、缓存和大模型权重。模型权重请放在独立的 checkpoint 仓库中管理。
cvlface/
research/
recognition/
code/
run_v1/
scripts/
train_plan4_adaface_swin.py
prepare_kaggle_verification_protocol.py
generate_plan8_report.py
configs/
*.yaml
test_*.py
其中 train_plan4_adaface_swin.py 是当前主要训练脚本。
--use-lora--lora-rank--lora-alpha--lora-dropout--lora-target-modules--eval-only--eval-checkpoint--eval-pairs-csv--config configs/xxx.yamltraining_config.jsonepoch_metrics.csv本代码仓库不包含数据集。当前项目中用到的数据备份在以下仓库:
lhx05/WebFace224lhx05/lab1-resized-face-datasetslhx05/fgnet-age30-protocollhx05/agedb-224-by-identity模型权重和训练 checkpoint 放在:
lhx05/lab1-face-recognition-checkpoints
本代码仓库不上传 .pt、.pth、.safetensors 等大权重文件。
全量微调示例:
cd /root/Lab1/project/cvlface/research/recognition/code/run_v1/scripts
source /root/Lab1/venv/bin/activate
python train_plan4_adaface_swin.py \
--train-root /root/Lab1/data/processed/agedb_harmonliu05_224_by_identity \
--output-dir /root/Lab1/experiments/agedb567_full_finetune_20epoch \
--ckpt /root/Lab1/project/cvlface/pretrained_models/model.safetensors \
--epochs 20 \
--batch-size 128 \
--eval-every-epochs 5 \
--eval-pairs-csv /root/Lab1/data/processed/fgnet_age30_protocol/facerec_val/fgnet_age30/pairs.csv
使用 YAML config 示例:
python train_plan4_adaface_swin.py \
--config configs/agedb567_lora_r4_fgnet_age30.yaml
research/recognition/code/run_v1/scripts/ 目录。