Commit 4635bdf7 authored by Jared Casper's avatar Jared Casper
Browse files

Merge branch 'readme_update_jan_2021' into 'main' 

Readme update + change gpt2 to gpt

See merge request ADLR/megatron-lm!206
parents 86eb5bd8 152aab30

README.md

+101 −124

File changed.

Preview size limit exceeded, changes collapsed.

examples/finetune_mnli_distributed.sh

+2 −2
Original line number Diff line number Diff line
@@ -28,11 +28,11 @@ python -m torch.distributed.launch $DISTRIBUTED_ARGS ./tasks/main.py \ 
               --num-layers 24 \

               --hidden-size 1024 \

               --num-attention-heads 16 \

               --batch-size 8 \

               --micro-batch-size 8 \

               --checkpoint-activations \

               --lr 5.0e-5 \

               --lr-decay-style linear \

               --warmup 0.065 \

               --lr-warmup-fraction 0.065 \

               --seq-length 512 \

               --max-position-embeddings 512 \

               --save-interval 500000 \

examples/finetune_race_distributed.sh

+2 −2
Original line number Diff line number Diff line
@@ -28,11 +28,11 @@ python -m torch.distributed.launch $DISTRIBUTED_ARGS ./tasks/main.py \ 
               --num-layers 24 \

               --hidden-size 1024 \

               --num-attention-heads 16 \

               --batch-size 4 \

               --micro-batch-size 4 \

               --checkpoint-activations \

               --lr 1.0e-5 \

               --lr-decay-style linear \

               --warmup 0.06 \

               --lr-warmup-fraction 0.06 \

               --seq-length 512 \

               --max-position-embeddings 512 \

               --save-interval 100000 \

examples/pretrain_bert.sh

+4 −4
Original line number Diff line number Diff line
@@ -9,24 +9,24 @@ python pretrain_bert.py \ 
       --num-layers 24 \

       --hidden-size 1024 \

       --num-attention-heads 16 \

       --batch-size 4 \

       --micro-batch-size 4 \

       --global-batch-size 8 \

       --seq-length 512 \

       --max-position-embeddings 512 \

       --train-iters 2000000 \

       --lr-decay-iters 990000 \

       --save $CHECKPOINT_PATH \

       --load $CHECKPOINT_PATH \

       --data-path $DATA_PATH \

       --vocab-file bert-vocab.txt \

       --data-impl mmap \

       --split 949,50,1 \

       --distributed-backend nccl \

       --lr 0.0001 \

       --min-lr 0.00001 \

       --lr-decay-style linear \

       --lr-decay-iters 990000 \

       --lr-warmup-fraction .01 \

       --weight-decay 1e-2 \

       --clip-grad 1.0 \

       --warmup .01 \

       --log-interval 100 \

       --save-interval 10000 \

       --eval-interval 1000 \