Edit model card
#!/bin/bash
# Apache v2 license
# Copyright (C) 2021 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

# Teacher Preparation

# Notes:
# Auto mixed precision can be used by adding --fp16
# Distributed training can be used with the torch.distributed.lauch app

TEACHER_PATH=./bert-base-uncased-teacher-preparation-pretrain
OUTPUT_DIR=$TEACHER_PATH
DATA_CACHE_DIR=/root/kaokao/Model-Compression-Research-Package/examples/transformers/language-modeling/wikipedia_processed_for_pretrain

python -m torch.distributed.launch \
    --nproc_per_node=8 \
    ../../examples/transformers/language-modeling/run_mlm.py \
    --model_name_or_path bert-base-uncased \
    --datasets_name_config wikipedia:20200501.en \
    --data_process_type segment_pair_nsp \
    --dataset_cache_dir $DATA_CACHE_DIR \
    --do_train \
    --learning_rate 5e-5 \
    --max_steps 100000 \
    --warmup_ratio 0.01 \
    --weight_decay 0.01 \
    --per_device_train_batch_size 8 \
    --gradient_accumulation_steps 4 \
    --logging_steps 10 \
    --save_steps 5000 \
    --save_total_limit 2 \
    --output_dir $OUTPUT_DIR \
    --run_name pofa-teacher-prepare-pretrain
Downloads last month
1
Hosted inference API

Unable to determine this model’s pipeline type. Check the docs .