Adding the benchmarking with TorchBench #5788
zpcore
merged
36 commits into
pytorch:benchmark
from
Liyang90:benchmark
Initial commit with dummy model benchmark
f2ed93b1
add XRT support
5984de6d
Add torchbench benchmark models
8d0f78e8
add randomize_input
ca20c6f8
add model set up for torchbench model
4ae79e25
update ExperimentLoader
35d446fa
Add saving results
4bb173da
minor args update
980cd5b3
sync with master
7cc45fb2
update style
a4106666
add experiment name
9e1506ed
add grad context for eval and train
7591cbce
minor user config update
6e9a3273
fix train() return item
1939fe08
minor refactor
8a50910d
add dynamo options
bacda048
add column in result for dynamo setting
0bc642ef
using to capture output and error
38511809
Fix some failure cases for dynamo
83c674ca
reduce eval result size by returning eval loss
113830f9
minor refactor
de609b81
revert eval result change
2d948361
minor fix
c2ad2781
Change output format to jsonl
88fee741
Add accelerator model nname
deb14825
sync with master
4993534f
add skipping finished experiments
10c52a7a
main process needs to remove PJRT_DEVICE env var that is automaticall…
3b3724c1
Add a simple result analyzer
668f2897
Result analyzer save to database csv with historical data
1e787a74
Handle detectron2 models
80a3fd62
Merge branch 'master' into benchmark
ad12d41b
minor update
a756875a
Merge branch 'master' into benchmark
aec61fbc
add deny list
307578c5
Merge branch 'pytorch:benchmark' into benchmark
84dff003
JackCaoG
approved these changes
on 2023-11-10
Login to write a write a comment.
Login via GitHub