文件列表:

.DS_Store (6148, 2022-05-22)
execute_sweep.sh (1325, 2022-05-22)
requirements.txt (17641, 2022-05-22)
setup.py (1101, 2022-05-22)
src/ (0, 2022-05-22)
src/.DS_Store (8196, 2022-05-22)
src/__init__.py (0, 2022-05-22)
src/data_processing/ (0, 2022-05-22)
src/data_processing/cache_tasks_main.py (7440, 2022-05-22)
src/data_processing/data_loading.py (1896, 2022-05-22)
src/data_processing/data_loading_tfds.py (10474, 2022-05-22)
src/data_processing/preprocessors.py (1840, 2022-05-22)
src/data_processing/setup.py (4785, 2022-05-22)
src/genformer.py (10111, 2022-05-22)
src/layers/ (0, 2022-05-22)
src/layers/__init__.py (0, 2022-05-22)
src/layers/fast_attention.py (21665, 2022-05-22)
src/layers/fast_attention_rpe.py (32677, 2022-05-22)
src/layers/fast_attention_rpe_genformer1.py (32501, 2022-05-22)
src/layers/layers.py (14263, 2022-05-22)
src/layers/util.py (8243, 2022-05-22)
src/losses.py (10784, 2022-05-22)
src/metrics.py (10997, 2022-05-22)
src/models/ (0, 2022-05-22)
src/models/.DS_Store (6148, 2022-05-22)
src/models/__init__.py (0, 2022-05-22)
src/models/genformer1.py (11889, 2022-05-22)
src/models/genformer_bn.py (0, 2022-05-22)
src/optimizers.py (14065, 2022-05-22)
src/schedulers.py (3166, 2022-05-22)
src/utils.py (2854, 2022-05-22)
src/version.py (21, 2022-05-22)
train_model.py (17681, 2022-05-22)
training_utils.py (27353, 2022-05-22)

# genformer dev repository **genformer** learns representations of accessible sequence via "masked accessibility modeling" which can be used for downstream fine-tuning tasks ## dataset inputs See https://app.terra.bio/#workspaces/epigenomics/gro_nn for data processing workflows and input data. ## main files For pre-training(masked atac prediction, _atac suffix files): * execute_sweep_atac.sh - training bash script where you can define hyperparameters * training_utils_atac.py - define train and validation steps, data loading and augmentation, masking, early stopping, model saving * train_model_atac.py - define main training loop, argument parsing, wandb initialization code, TPU initialization code * src/models/aformer_atac.py - main model file * src/layers/layers.py - all custom layers * src/layers/fast_attention_rpe_genformer1.py - linear attention code with rotary positional encodings Files for fine-tuning for RAMPAGE prediction follow a similar structure ## training Define hyper- and sweep parameters in execute_sweep_atac.sh

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：