From dc801c07cf38b0c495686463e6ca6f871a64440e Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Tue, 27 Jan 2026 09:57:37 -0600 Subject: Add collaborativeagents module and update gitignore - Add collaborativeagents subproject with adapters, agents, and evaluation modules - Update .gitignore to exclude large binary files (.whl, .tar), wandb logs, and results Co-Authored-By: Claude Opus 4.5 --- collaborativeagents/training/train_sft_lf.sbatch | 40 ++++++++++++++++++++++++ 1 file changed, 40 insertions(+) create mode 100644 collaborativeagents/training/train_sft_lf.sbatch (limited to 'collaborativeagents/training/train_sft_lf.sbatch') diff --git a/collaborativeagents/training/train_sft_lf.sbatch b/collaborativeagents/training/train_sft_lf.sbatch new file mode 100644 index 0000000..7f090b1 --- /dev/null +++ b/collaborativeagents/training/train_sft_lf.sbatch @@ -0,0 +1,40 @@ +#!/bin/bash +#SBATCH --job-name=sft_lf +#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=8:00:00 +#SBATCH --output=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_lf_%j.out +#SBATCH --error=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_lf_%j.err + +echo "=== SFT Training with LLaMA-Factory (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: /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/models/llama-3.1-8b-instruct" +echo "Data: training_data/sft_training_data.json" +echo "Output: outputs/sft_reflection_lf" + +# Check training data +echo "Training data size: $(stat -c%s training_data/sft_training_data.json) bytes" +echo "Training examples: $(python3 -c 'import json; print(len(json.load(open(\"training_data/sft_training_data.json\"))))')" + +echo "" +echo "Starting LLaMA-Factory SFT training with DeepSpeed ZeRO-3..." + +# Run with LLaMA-Factory CLI + DeepSpeed ZeRO-3 for memory efficiency +FORCE_TORCHRUN=1 llamafactory-cli train llama_factory_config.yaml + +echo "" +echo "SFT Training complete!" +date -- cgit v1.2.3