Files
ascend-tools/train/yaml/Bert-Base.yaml
T
2020-10-19 20:22:23 +08:00

47 lines
1.5 KiB
YAML

tensorflow_config:
#layer层数有6和12两种,中文数据集用 bert_base_layer6_cn.json/bert_base_layer12_cn.json 英文用bert_base_layer6_cn.json/bert_base_layer12_en.json
bert_config_file: bert_base_layer6_cn.json
#数据集句子长度是256时 设置为 256,40,句子长度是128时设置为128,20
max_seq_length: 128
max_predictions_per_seq: 20
# 最佳性能train_batch_size为160
train_batch_size: 160
learning_rate: 1e-4
num_warmup_steps: 100
num_train_steps: 1000
optimizer_type: adam
manual_fp16: True
use_fp16_cls: True
input_files_dir: /home/BertData/cn-wiki-128/
eval_files_dir: /home/BertData/cn-wiki-128/
npu_bert_debug: False
npu_bert_use_tdt: True
distributed: True
do_train: True
do_eval: False
num_accumulation_steps: 1
iterations_per_loop: 100
npu_bert_loss_scale: 0
save_checkpoints_steps: 1000
npu_bert_clip_by_global_norm: False
# docker 镜像名称:版本号
docker_image: c73:b021
# 仅多机执行需要配置: ip1:卡数量1,ip2:卡数量2
mpirun_ip: 90.90.140.199:8,90.90.140.229:8
# 指定 device id, 多个 id 使用空格分隔, 数量需与 rank_size 相同
device_group_1p: 6
device_group_2p: 0 1
device_group_4p: 0 1 2 3
#profiling 配置
PROFILING_MODE: false
AICPU_PROFILING_MOD: false
PROFILING_OPTIONS: training_trace
FP_POINT: bert/embeddings/GatherV2
BP_POINT: gradients/bert/embeddings/IdentityN_1_grad/UnsortedSegmentSum