提交 16a8ac54 编写于 作者: T tangwei

fix cluster

上级 08e791fa
......@@ -14,20 +14,13 @@
train:
trainer:
#trainer: "fleetrec/demo/user_define_trainer.py"
threads: 4
# for cluster training
strategy: "async"
communicator:
send_queue_size: 4
min_send_grad_num_before_recv: 4
thread_pool_size: 5
max_merge_var_num: 4
epochs: 10
reader:
mode: "dataset"
batch_size: 2
class: "fleetrec.models.rank.criteo_reader"
train_data_path: "models/rank/dnn/data/train"
......
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