DeepLearningExamples/PyTorch/SpeechSynthesis/Tacotron2/qa/L2_Tacotron2_inferbench_FP16_DGX1_16GB_1GPU_single.sh

18 lines
549 B
Bash

#!/bin/bash
set -e
## uncomment to generate new baseline; will be created in qa/baselines/ ##
## python inference_perf.py -m Tacotron2 -bs=20 --input-text qa/text_padded.pt --create-benchmark
python inference_perf.py -m Tacotron2 -bs=1 --fp16-run --decoder-no-early-stopping --input-text qa/text_padded.pt
python qa/check_curves.py \
-b qa/tacotron2_fp16-infer-bs1.json \
Tacotron2_infer_BS1_FP16_DGX1_16GB_1GPU_single.json \
-g "iter" \
-k "items_per_sec" \
--eps 0.001 \
--damping 1 \
--sigma 12.0 \
--fail low