A Deep Learning Pipeline
Project description
dl_pipeline
A general deep learning pipeline (in construction) for kaggle competitions and other projects.
Install
Setup with pip:
pip install dl_pipeline
Clone and editable setup:
git clone https://github.com/mnpinto/dl_pipeline
cd dl_pipeline
pip install -e .
Rainforest Connection Species Audio Detection
#!/bin/bash
arch='densenet121'
model_name='model_0'
sample_rate=32000
n_mels=128
hop_length=640
for fold in 0 1 2 3 4
do
echo "Training $model for fold $fold"
kaggle_rainforest2021 --fold $fold --model_name $model_name \
--model $arch --sample_rate $sample_rate --n_mels $n_mels \
--hop_length $hop_length --bs 32 --head_ps 0.8 \
--tile_width 1024 --mixup true >> log.train
done
for tw in 64 128 256
do
echo "Generate predictions for $model with tile_width of $tw"
kaggle_rainforest2021 --run_test true --model_name $model_name \
--model $arch --sample_rate $sample_rate --n_mels $n_mels \
--hop_length $hop_length --tile_width $tw \
--save_preds true >> log.predict
done
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
dl_pipeline-0.0.1.tar.gz
(26.3 kB
view hashes)
Built Distribution
Close
Hashes for dl_pipeline-0.0.1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 58a692eb5494714a74525e489c9947e9b8ecf8cc7934fb531243a28d4b4bf86e |
|
MD5 | f65fdb2c34230b77fe2786f96a851e26 |
|
BLAKE2b-256 | d5c583af922b59d9476227e0b8fc2335f92307b78ad5364b19b4ef3bbe010c43 |