File size: 1,096 Bytes
62dca4c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 | #!/bin/bash
SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
ROOT_DIR=$(dirname $SCRIPT_DIR)
export TORCHINDUCTOR_CACHE_DIR=$ROOT_DIR/cache/compiled_kernels
NUM_GPUS=${1:-1}
TP_SIZE=${2:-1}
BUILD_DATASET_NUM_PROC=${BUILD_DATASET_NUM_PROC:-64}
torchrun \
--standalone \
--nproc_per_node $NUM_GPUS \
$ROOT_DIR/scripts/train_eagle3.py \
--target-model-path meituan-longcat/LongCat-Flash-Chat-FP8 \
--draft-model-config $ROOT_DIR/configs/longcat-flash-eagle3.json \
--train-data-path $ROOT_DIR/cache/dataset/sharegpt_train.jsonl \
--build-dataset-num-proc $BUILD_DATASET_NUM_PROC \
--output-dir $ROOT_DIR/outputs/longcat-flash-eagle3-sharegpt \
--num-epochs 10 \
--batch-size 1 \
--tp-size $TP_SIZE \
--learning-rate 1e-4 \
--max-length 2048 \
--chat-template longcat \
--cache-dir $ROOT_DIR/cache \
--attention-backend sdpa \
--target-model-backend sglang \
--log-interval 10 \
--sglang-mem-fraction-static 0.75 \
--sglang-attention-backend flashinfer \
--sglang-ep-size $NUM_GPUS
|