From dc801c07cf38b0c495686463e6ca6f871a64440e Mon Sep 17 00:00:00 2001
From: YurenHao0426 <blackhao0426@gmail.com>
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 <noreply@anthropic.com>
---
 collaborativeagents/slurm/run_sft_only.sh | 64 +++++++++++++++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 collaborativeagents/slurm/run_sft_only.sh

(limited to 'collaborativeagents/slurm/run_sft_only.sh')

diff --git a/collaborativeagents/slurm/run_sft_only.sh b/collaborativeagents/slurm/run_sft_only.sh
new file mode 100644
index 0000000..a98ae25
--- /dev/null
+++ b/collaborativeagents/slurm/run_sft_only.sh
@@ -0,0 +1,64 @@
+#!/bin/bash
+#SBATCH --job-name=sft_train
+#SBATCH --account=bfqt-delta-gpu
+#SBATCH --partition=gpuA100x4
+#SBATCH --nodes=1
+#SBATCH --ntasks=1
+#SBATCH --cpus-per-task=16
+#SBATCH --gres=gpu:4
+#SBATCH --mem=200G
+#SBATCH --time=24:00:00
+#SBATCH --output=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_train_%j.out
+#SBATCH --error=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/sft_train_%j.err
+
+# SFT Training only (data already exists)
+
+set -e
+cd /projects/bfqt/users/yurenh2/ml-projects/personalization-user-model
+
+source /u/yurenh2/miniforge3/etc/profile.d/conda.sh
+conda activate eval
+
+export HF_HOME=/projects/bfqt/users/yurenh2/hf_cache/huggingface
+export PYTHONPATH="${PWD}/src:${PWD}/collaborativeagents:${PWD}/collaborativeagents/scripts:${PYTHONPATH}"
+export WANDB_PROJECT="collaborative-agent-reflection-sft"
+
+echo "=== SFT Training ==="
+date
+nvidia-smi --query-gpu=index,name,memory.total --format=csv
+
+DATA_PATH="collaborativeagents/training/training_data/sft_training_data.json"
+MODEL_PATH="/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/models/llama-3.1-8b-instruct"
+OUTPUT_DIR="collaborativeagents/training/outputs/sft_reflection"
+
+echo "Data: $DATA_PATH"
+echo "Model: $MODEL_PATH"
+echo "Output: $OUTPUT_DIR"
+
+# Check data exists
+if [ ! -f "$DATA_PATH" ]; then
+    echo "ERROR: Training data not found at $DATA_PATH"
+    exit 1
+fi
+
+echo ""
+echo "Training data size: $(wc -c < $DATA_PATH) bytes"
+python -c "import json; d=json.load(open('$DATA_PATH')); print(f'Training examples: {len(d)}')"
+
+mkdir -p $OUTPUT_DIR
+
+echo ""
+echo "Starting SFT training..."
+python collaborativeagents/training/train_sft.py \
+    --model-path $MODEL_PATH \
+    --data-path $DATA_PATH \
+    --output-dir $OUTPUT_DIR \
+    --num-epochs 4 \
+    --learning-rate 1e-6 \
+    --batch-size 1 \
+    --gradient-accumulation 64
+
+echo ""
+echo "=== SFT Training Complete ==="
+echo "Model saved to: $OUTPUT_DIR"
+date
-- 
cgit v1.2.3