1 files changed, 29 insertions, 0 deletions
diff --git a/collaborativeagents/training/run_train.sh b/collaborativeagents/training/run_train.sh
new file mode 100644
index 0000000..a7f8595
--- /dev/null
+++ b/collaborativeagents/training/run_train.sh
@@ -0,0 +1,29 @@
+# vllm serve meta-llama/Llama-3.3-70B-Instruct --port 8004 --tensor-parallel-size 4 --max-model-len 16384 --gpu-memory-utilization 0.9
+# python -m sglang.launch_server --model-path meta-llama/Llama-3.3-70B-Instruct --port 8004 --tp-size 4 --context-length 16384
+# python -m sglang.launch_server --model-path meta-llama/Llama-3.1-8B-Instruct --port 8003 --tp-size 4 --context-length 16384
+
+
+BATCH_SIZE=100
+BATCH_SIZE=50
+
+# Loop over eval sizes and datasets
+for EVAL_SIZE in 20; do
+    for DATASET in math-hard math-500 logiqa mmlu medqa; do # humaneval bigcodebench
+        # Convert dataset name for file paths (replace - with _)
+        DATASET_FILE=$(echo ${DATASET} | tr '-' '_')
+
+        echo "Running experiments for dataset: ${DATASET} with eval_size ${EVAL_SIZE}"
+
+        # training_data_with_user_profiles_with_preferences
+        python3 run.py --experiment_type training_data_with_user_profiles_with_preferences --dataset ${DATASET} --eval_size ${EVAL_SIZE} --max_turns 10 --batch_size ${BATCH_SIZE} \
+            --user_model_name hosted_vllm/meta-llama/Llama-3.3-70B-Instruct --user_api_base http://localhost:8004/v1 --user_api_key EMPTY \
+            --collaborator_model_name hosted_vllm/meta-llama/Llama-3.3-70B-Instruct --collaborator_api_base http://localhost:8004/v1 --collaborator_api_key EMPTY \
+            --judge_model_name hosted_vllm/meta-llama/Llama-3.3-70B-Instruct --judge_api_base http://localhost:8004/v1 --judge_api_key EMPTY \
+            --output_file /shared/storage-01/users/mehri2/mem/collaborativeagents/scripts/runs/training_llama70b_temp_1/${DATASET_FILE}_llama70b_user_llama70b_agent_training_data_with_reflection_eval_size_${EVAL_SIZE}.jsonl \
+            >> ./runs/training_llama70b_temp_1/${DATASET_FILE}_llama70b_user_llama70b_agent_training_data_with_reflection_eval_size_${EVAL_SIZE}.out 2>&1
+
+    done
+done
+
+
+