forked from tenstorrent/benchmarking
-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_benchmark_cuda
58 lines (39 loc) · 2.9 KB
/
run_benchmark_cuda
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
#!/bin/bash
# Start benchmarking
echo "Running TT-Model Benchmarking on CUDA"
# BERT - Text Classification - SST-2
python benchmark.py -d cuda -m bert -c large --task text_classification -df Fp16 -mb 64 --loop_count 4 --save_output
# ResNet - Image Classification - ImageNet
python benchmark.py -d cuda -m resnet -c resnet50 --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# T5 - Text Summarization - CNN Dailymail
python benchmark.py -d cuda -m t5 -c large --task text_summarization -df Fp16 -mb 1 --loop_count 1 --save_output
# FLAN-T5 - Text Summarization - CNN Dailymail
python benchmark.py -d cuda -m flant5 -c large --task text_summarization -df Fp16 -mb 1 --loop_count 1 --save_output
# VoVNet - Image Classification - ImageNet
python benchmark.py -d cuda -m vovnet_v2 -c 39 --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# MobileNetV1 - Image Classification - ImageNet
python benchmark.py -d cuda -m mobilenetv1 -c 224 --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# MobileNetV2 - Image Classification - ImageNet
python benchmark.py -d cuda -m mobilenetv2 -c 224 --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# MobileNetV3 - Image Classification - ImageNet
python benchmark.py -d cuda -m mobilenetv3 -c lg --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# HRNet - Image Classification - ImageNet
python benchmark.py -d cuda -m hrnet -c v2_w64 --task image_classification -df Fp16 -mb 64 --loop_count 4 --save_output
# ViT - Image Classification - ImageNet
python benchmark.py -d cuda -m vit -c base --task image_classification -df Fp16 -mb 64 --loop_count 4 --save_output
# DeiT - Image Classification - ImageNet
python benchmark.py -d cuda -m deit -c base --task image_classification -df Fp16 -mb 64 --loop_count 4 --save_output
# YOLOv5 - Object Detecton - COCO
python benchmark.py -d cuda -m yolo_v5 -c s --task object_detection -df Fp16 -mb 64 --loop_count 1 --save_output
# OpenPose - Pose Estimation - COCO keypoints
python benchmark.py -d cuda -m open_pose -c 2d --task pose_estimation -mb 64 -df Fp16 --loop_count 4 --save_output
# U-Net - Segmentation - Brain LGG Dataset
python benchmark.py -d cuda -m unet -c 256 --task segmentation -df Fp16 -mb 48 --loop_count 1 --save_output
# Inception-v4 - Image Classification - ImageNet
python benchmark.py -d cuda -m inception_v4 -c 224 --task image_classification -df Fp16 -mb 64 --loop_count 8 --save_output
# Whisper - ASR - LibriSpeech
python benchmark.py -d cuda -m whisper -c small --task asr -df Fp32 -mb 1 --loop_count 1 --save_output
# Falcon - Text Generation - HellaSwag
python benchmark.py -d cuda -m falcon -c 7b-instruct --task hellaswag -df Fp16 -mb 32 --loop_count 1 --save_output
# Stable Diffusion - Image Generation
python benchmark.py -d cuda -m stable_diffusion -c v1-4 --task image_generation -df Fp16 -mb 1 --loop_count 1 --save_output