@tjysdsg @VincieSlytherin @kurisujhin @SandyLuXY
- Init all git submodules
- Enter
pretrained/
and pull git lfs filesgit lfs install git lfs pull
- Install ESPnet: https://espnet.github.io/espnet/installation.html
- Prepare data
cd must_c_test_subset python ../prepare_data.py --out_dir ../data --sample_rate 16000
- Pretrained model that's used: https://huggingface.co/espnet/brianyan918_mustc-v2_en-de_st_conformer_asrinit_v2_raw_en_de_bpe_tc4000_sp
cd pretrained
python plot_lab2.py --result_dir output --out_dir output/plot_all
lab3.py
plot_all.py
lab4.py
lab4_export_onnx.py
lab4_benchmark_onnx.py
-
Export ONNX models w/o other techniques
cd pretrained/ python export_onnx.py --pruned --quantized
-
Run
run_lab5.sh
andrun_lab5_gpu.sh
. Check comments carefully before running.
Under pretrained/
:
-
Export ONNX models w/o other techniques
cd pretrained/ python ../export_onnx.py --pruned --quantized # adjust these flags as needed
-
Run benchmarks
cd pretrained/ python ../final_benchmark.py --data_dir ../data --out_dir ../output
output_onnx/
is constructed from results of final_benchmark.py
, containing ablation study of ONNX, pruning, and
quantization.
Run plot_all.py
on this folder to generate the plot.
- ESPnet: https://github.com/espnet/espnet
- ESPnet documentation: https://espnet.github.io/espnet/
- https://github.com/mjpost/sacrebleu
- Source code for speech translation inference: https://github.com/espnet/espnet/blob/master/espnet2/bin/st_inference.py
- Source code for speech translation model: https://github.com/espnet/espnet/blob/master/espnet2/st/espnet_model.py