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 --- .../training/train_sft_qlora.sbatch | 29 ++++++++++++++++++++++ 1 file changed, 29 insertions(+) create mode 100644 collaborativeagents/training/train_sft_qlora.sbatch (limited to 'collaborativeagents/training/train_sft_qlora.sbatch') diff --git a/collaborativeagents/training/train_sft_qlora.sbatch b/collaborativeagents/training/train_sft_qlora.sbatch new file mode 100644 index 0000000..81c3db1 --- /dev/null +++ b/collaborativeagents/training/train_sft_qlora.sbatch @@ -0,0 +1,29 @@ +#!/bin/bash +#SBATCH --job-name=sft_qlora +#SBATCH --output=sft_qlora_%j.out +#SBATCH --error=sft_qlora_%j.err +#SBATCH --time=12:00:00 +#SBATCH --partition=a100 +#SBATCH --gres=gpu:1 +#SBATCH --cpus-per-task=8 +#SBATCH --mem=64G + +# QLoRA SFT Training - Ultra Minimal Hardware (1 GPU, ~12GB VRAM) +# Can run on A100 40GB, V100, or even consumer GPUs + +cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training + +source /u/yurenh2/miniforge3/etc/profile.d/conda.sh +conda activate eval + +echo "=== Starting QLoRA SFT Training ===" +echo "GPU: $(nvidia-smi --query-gpu=name --format=csv,noheader | head -1)" +echo "Config: llama_factory_qlora_config.yaml" +echo "Start time: $(date)" + +# Run training with LLaMA-Factory +llamafactory-cli train llama_factory_qlora_config.yaml + +echo "=== Training Complete ===" +echo "End time: $(date)" +echo "Output: outputs/sft_reflection_qlora" -- cgit v1.2.3