#!/bin/bash set -e CURRENT_DIR="$( cd "$( dirname "$0" )" && pwd )" MEGATRON_PATH=$( dirname $( dirname ${CURRENT_DIR})) export PYTHONPATH=$PYTHONPATH:${MEGATRON_PATH}:${MEGATRON_PATH}/PAI-Megatron-LM-240718 export CUDA_DEVICE_MAX_CONNECTIONS=1 ENV=$1 if [ $ENV = dsw ]; then export CUDA_VISIBLE_DEVICES=7 MASTER_ADDR=localhost MASTER_PORT=$(shuf -n 1 -i 10000-65535) NNODES=1 NODE_RANK=0 GPUS_PER_NODE=1 elif [ $ENV = dlc ]; then NNODES=${WORLD_SIZE} NODE_RANK=${RANK} GPUS_PER_NODE=${KUBERNETES_CONTAINER_RESOURCE_GPU} fi DISTRIBUTED_ARGS="--nproc_per_node $GPUS_PER_NODE --nnodes $NNODES --node_rank $NODE_RANK --master_addr $MASTER_ADDR --master_port $MASTER_PORT" MODEL_SIZE=$2 BATCH_SIZE=$3 GLOBAL_BATCH_SIZE=$4 LR=$5 MIN_LR=$6 SEQ_LEN=$7 PAD_LEN=$8 PR=$9 TP=${10} PP=${11} EP=${12} AC=${13} DO=${14} FL=${15} SP=${16} TE=${17} SAVE_INTERVAL=${18} DATASET_PATH=${19} PRETRAIN_CHECKPOINT_PATH=${20} TRAIN_TOKENS=${21} WARMUP_TOKENS=${22} OUTPUT_BASEPATH=${23} if [ $MODEL_SIZE = 0.5B ]; then HIDDEN_SIZE=896 INTERMEDIATE_SIZE=4864 MAX_POSITION_EMBEDDINGS=131072 MAX_WINDOW_LAYERS=24 NUM_ATTENTION_HEADS=14 NUM_HIDDEN_LAYERS=24 NUM_KEY_VALUE_HEADS=2 RMS_NORM_EPS=1e-6 ROPE_THETA=1000000 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=34 moe_options=" \ " elif [ $MODEL_SIZE = 1.5B ]; then HIDDEN_SIZE=1536 INTERMEDIATE_SIZE=8960 MAX_POSITION_EMBEDDINGS=131072 MAX_WINDOW_LAYERS=28 NUM_ATTENTION_HEADS=12 NUM_HIDDEN_LAYERS=28 NUM_KEY_VALUE_HEADS=2 RMS_NORM_EPS=1e-6 ROPE_THETA=1000000 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=293 moe_options=" \ " elif [ $MODEL_SIZE = 7B ]; then HIDDEN_SIZE=3584 INTERMEDIATE_SIZE=18944 MAX_POSITION_EMBEDDINGS=131072 MAX_WINDOW_LAYERS=28 NUM_ATTENTION_HEADS=28 NUM_HIDDEN_LAYERS=28 NUM_KEY_VALUE_HEADS=4 RMS_NORM_EPS=1e-6 ROPE_THETA=1000000 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=421 moe_options=" \ " elif [ $MODEL_SIZE = 70B ]; then HIDDEN_SIZE=8192 INTERMEDIATE_SIZE=28672 MAX_POSITION_EMBEDDINGS=4096 MAX_WINDOW_LAYERS=80 NUM_ATTENTION_HEADS=64 NUM_HIDDEN_LAYERS=80 NUM_KEY_VALUE_HEADS=8 RMS_NORM_EPS=1e-5 ROPE_THETA=1000000 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=0 moe_options=" \ " elif [ $MODEL_SIZE = 72B ]; then HIDDEN_SIZE=8192 INTERMEDIATE_SIZE=29568 MAX_POSITION_EMBEDDINGS=131072 MAX_WINDOW_LAYERS=80 NUM_ATTENTION_HEADS=64 NUM_HIDDEN_LAYERS=80 NUM_KEY_VALUE_HEADS=8 RMS_NORM_EPS=1e-5 ROPE_THETA=1000000 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=421 moe_options=" \ " elif [ $MODEL_SIZE = A14B ]; then HIDDEN_SIZE=3584 INTERMEDIATE_SIZE=18944 MAX_POSITION_EMBEDDINGS=131072 MAX_WINDOW_LAYERS=28 MOE_INTERMEDIATE_SIZE=2560 NUM_ATTENTION_HEADS=28 NUM_EXPERTS=64 NUM_EXPERTS_PER_TOPK=8 NUM_HIDDEN_LAYERS=28 NUM_KEY_VALUE_HEADS=4 RMS_NORM_EPS=1e-6 ROPE_THETA=1000000 SHARED_EXPERT_INTERMEDIATE_SIZE=20480 SLIDING_WINDOW=131072 EXTRA_VOCAB_SIZE=293 moe_options=" \ --moe-router-topk ${NUM_EXPERTS_PER_TOPK} \ --num-experts ${NUM_EXPERTS} \ --expert-model-parallel-size ${EP}\ --moe-ffn-hidden-size ${MOE_INTERMEDIATE_SIZE} \ --shared-moe-ffn-hidden-size ${SHARED_EXPERT_INTERMEDIATE_SIZE} \ --enable-shared-expert" fi if [ $AC = full ]; then activation_checkpoint_options=" \ --recompute-method uniform \ --recompute-num-layers 1 \ --recompute-granularity full" elif [ $AC = sel ]; then activation_checkpoint_options=" \ --recompute-activations" elif [ $AC = none ]; then activation_checkpoint_options=" \ " fi if [ $PR = fp16 ]; then pr_options=" \ --fp16 \ --apply-query-key-layer-scaling" export NVTE_APPLY_QK_LAYER_SCALING=1 elif [ $PR = bf16 ]; then pr_options=" \ --bf16" elif [ $PR = fp8 ]; then pr_options=" \ --bf16 \ --fp8-hybrid \ --fp8-amax-compute-algo max \ --fp8-amax-history-len 1024 \ --transformer-impl transformer_engine" fi if [ $DO = true ]; then do_options=" \ --use-distributed-optimizer" elif [ $DO = false ]; then do_options=" \ " fi if [ $FL = true ]; then flash_options=" \ --use-flash-attn" elif [ $FL = false ]; then flash_options=" \ " fi if [ $TE = true ]; then te_options=" \ --transformer-impl transformer_engine" elif [ $TE = false ]; then te_options=" \ --transformer-impl local" fi if [ $SP = true ] && [ $TP -gt 1 ]; then sp_options=" \ --sequence-parallel" elif [ $SP = false ]; then sp_options=" \ " fi if [ $PRETRAIN_CHECKPOINT_PATH != none ]; then load_options=" \ --load $PRETRAIN_CHECKPOINT_PATH" fi TRAIN_ITERS=$(( ${TRAIN_TOKENS} / ${GLOBAL_BATCH_SIZE} / ${SEQ_LEN} )) LR_WARMUP_ITERS=$(( ${WARMUP_TOKENS} / ${GLOBAL_BATCH_SIZE} / ${SEQ_LEN} )) LR_DECAY_ITERS=$(( ${TRAIN_TOKENS} / ${GLOBAL_BATCH_SIZE} / ${SEQ_LEN} )) NAME="pretrain-mcore-llama3-${MODEL_SIZE}-lr-${LR}-minlr-${MIN_LR}-bs-${BATCH_SIZE}-gbs-${GLOBAL_BATCH_SIZE}-seqlen-${SEQ_LEN}-pr-${PR}-tp-${TP}-pp-${PP}-ac-${AC}-do-${DO}-sp-${SP}-tt-${TRAIN_TOKENS}-wt-${WARMUP_TOKENS}" mkdir -p "${OUTPUT_BASEPATH}/tensorboard/" mkdir -p "${OUTPUT_BASEPATH}/checkpoint/" mkdir -p "${OUTPUT_BASEPATH}/log/" current_time=$(date "+%Y.%m.%d-%H.%M.%S") TENSORBOARD_DIR="${OUTPUT_BASEPATH}/tensorboard/${NAME}_${current_time}" mkdir -p ${TENSORBOARD_DIR} SAVED_PRETRAIN_CHECKPOINT_PATH="${OUTPUT_BASEPATH}/checkpoint/${NAME}" megatron_options=" \ --save ${SAVED_PRETRAIN_CHECKPOINT_PATH} \ --data-path ${DATASET_PATH} \ --split 99,1,0 \ --lr ${LR} \ --min-lr ${MIN_LR} \ --lr-decay-style cosine \ --weight-decay 0.1 \ --adam-beta1 0.9 \ --adam-beta2 0.95 \ --clip-grad 0.0 \ --init-method-std 0.008 \ --attention-dropout 0.0 \ --hidden-dropout 0.0 \ --lr-decay-iters ${LR_DECAY_ITERS} \ --lr-warmup-iters ${LR_WARMUP_ITERS} \ --train-iters ${TRAIN_ITERS} \ --micro-batch-size ${BATCH_SIZE} \ --global-batch-size ${GLOBAL_BATCH_SIZE} \ --num-layers ${NUM_HIDDEN_LAYERS} \ --hidden-size ${HIDDEN_SIZE} \ --num-attention-heads ${NUM_ATTENTION_HEADS} \ --ffn-hidden-size ${INTERMEDIATE_SIZE} \ --seq-length ${SEQ_LEN} \ --max-position-embeddings ${MAX_POSITION_EMBEDDINGS} \ --max-padding-length ${PAD_LEN} \ --log-interval 1 \ --eval-interval 10000 \ --eval-iters 10 \ --save-interval ${SAVE_INTERVAL} \ --tensorboard-queue-size 1 \ --tensorboard-dir ${TENSORBOARD_DIR} \ --log-timers-to-tensorboard \ --log-batch-size-to-tensorboard \ --log-validation-ppl-to-tensorboard \ --tensor-model-parallel-size ${TP} \ --pipeline-model-parallel-size ${PP} \ --no-load-optim \ --no-load-rng \ --num-workers 8 \ --extra-vocab-size ${EXTRA_VOCAB_SIZE} \ --patch-tokenizer-type LLama3Tokenizer \ --dataset LLama-Pretrain-Idxmap \ --swiglu \ --normalization RMSNorm \ --norm-epsilon ${RMS_NORM_EPS} \ --use-rotary-position-embeddings \ --no-rope-fusion \ --position-embedding-type rope \ --untie-embeddings-and-output-weights \ --no-log-loss-scale-to-tensorboard \ --disable-bias-linear \ --group-query-attention \ --num-query-groups ${NUM_KEY_VALUE_HEADS} \ --rotary-percent 1.0 \ --rotary-base ${ROPE_THETA} \ --rotary-seq-len-interpolation-factor 1 \ --optimizer cpu-adam " run_cmd="torchrun $DISTRIBUTED_ARGS pretrain_llama_mcore070.py ${megatron_options} ${pr_options} ${load_options} ${te_options} ${activation_checkpoint_options} ${do_options} ${flash_options} ${sp_options} ${moe_options}" echo ${run_cmd} eval ${run_cmd} set +x