Add Benchmark Script for Bert Models #3829
Add benchmark script for BERT models
c3e33aaf
Add license
37a04931
Set intra_op_num_threads=1 for cpu (version <= 1.2.0)
add1de35
Use comma instead of space to split models string
66e2ef7b
format
ad2909cf
Add percentiles for latency
10dd103d
Constraint models to predefined list.
44426c1a
torch.set_num_threads (for intra op) to get fair comparison
4fef83e9
Allow export ONNX model with specified number of inputs
70b7093c
Add warning that--num_inputs is not implemented for torch/torchscript
02d69397
change default test_times to 1000
31144a98
Add --optimize_onnx and --num_inputs options
514d592e
rename file
bf8a3f24
Rename file to benchmark.py
e53f6b60
Output a summary file
cbbe3a06
update comments
e9d9d945
Add xlnet-base-cased
936aeb2c
Add openai-gpt
30dc0945
Update default batch_sizes and sequence_lengths
7c73bb96
Install transformers from source
d9586448
liuziyue
approved these changes
on 2020-05-12
tianleiwu
merged
29496170
into master 5 years ago
tianleiwu
deleted the tlwu/benchmark_bert branch 5 years ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub