blob: e6a868b5198c27cf84492a572abfabc9c284a195 (
plain)
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
34
35
36
37
38
39
40
41
42
43
44
|
#!/bin/bash
#SBATCH --job-name=sft_resume
#SBATCH --account=bfqt-delta-gpu
#SBATCH --partition=gpuH200x8
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=32
#SBATCH --gres=gpu:4
#SBATCH --mem=256G
#SBATCH --time=1:00:00
#SBATCH --output=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_resume_%j.out
#SBATCH --error=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_resume_%j.err
echo "=== SFT Training Resume from checkpoint-100 (H200) ==="
date
nvidia-smi --query-gpu=index,name,memory.total --format=csv
cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training
source /u/yurenh2/miniforge3/etc/profile.d/conda.sh
conda activate eval
export HF_HOME=/projects/bfqt/users/yurenh2/hf_cache/huggingface
export WANDB_MODE=offline
echo "Model: llama-3.1-8b-instruct"
echo "Resuming from: /work/nvme/bfqt/yurenh2/sft_checkpoints/checkpoint-100"
echo "Output: /work/nvme/bfqt/yurenh2/sft_checkpoints"
# Verify checkpoint exists
if [ -d "/work/nvme/bfqt/yurenh2/sft_checkpoints/checkpoint-100" ]; then
echo "Checkpoint verified!"
else
echo "ERROR: Checkpoint not found!"
exit 1
fi
echo ""
echo "Starting LLaMA-Factory SFT training resume with DeepSpeed ZeRO-3..."
# Run with LLaMA-Factory CLI + DeepSpeed ZeRO-3
FORCE_TORCHRUN=1 llamafactory-cli train llama_factory_resume_config.yaml
echo ""
echo "SFT Training complete!"
date
|
