提交 f469dfe8 编写于 作者: G gaotingquan 提交者: cuicheng01

decrease bs

上级 9d92f768
......@@ -43,7 +43,7 @@ Optimizer:
momentum: 0.9
lr:
name: Cosine
learning_rate: 7.8
learning_rate: 5.2
warmup_epoch: 5
regularizer:
name: 'L2'
......@@ -73,11 +73,11 @@ DataLoader:
sampler:
name: DistributedBatchSampler
batch_size: 1536
batch_size: 1024
drop_last: False
shuffle: True
loader:
num_workers: 20
num_workers: 16
use_shared_memory: True
Eval:
......
......@@ -50,11 +50,11 @@ inference:python/predict_cls.py -c configs/inference_cls.yaml
-o Global.benchmark:False
null:null
===========================train_benchmark_params==========================
batch_size:1536
batch_size:1024
fp_items:fp16
epoch:1
model_type:norm_train
num_workers:20
num_workers:16
--profiler_options:batch_range=[10,20];state=GPU;tracer_option=Default;profile_path=model.profile
flags:FLAGS_eager_delete_tensor_gb=0.0;FLAGS_fraction_of_gpu_memory_to_use=0.98;FLAGS_conv_workspace_size_limit=4096
===========================infer_benchmark_params==========================
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册