summaryrefslogtreecommitdiff
path: root/collaborativeagents/training/train_sft_resume.sbatch
diff options
context:
space:
mode:
authorYurenHao0426 <blackhao0426@gmail.com>2026-01-27 09:57:37 -0600
committerYurenHao0426 <blackhao0426@gmail.com>2026-01-27 09:57:37 -0600
commitdc801c07cf38b0c495686463e6ca6f871a64440e (patch)
tree599f03114775921dbc472403c701f4a3a8ea188a /collaborativeagents/training/train_sft_resume.sbatch
parente43b3f8aa36c198b95c1e46bea2eaf3893b13dc3 (diff)
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>
Diffstat (limited to 'collaborativeagents/training/train_sft_resume.sbatch')
-rw-r--r--collaborativeagents/training/train_sft_resume.sbatch44
1 files changed, 44 insertions, 0 deletions
diff --git a/collaborativeagents/training/train_sft_resume.sbatch b/collaborativeagents/training/train_sft_resume.sbatch
new file mode 100644
index 0000000..e6a868b
--- /dev/null
+++ b/collaborativeagents/training/train_sft_resume.sbatch
@@ -0,0 +1,44 @@
+#!/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