Views
No views yet
parent_dir/
├── Q-RAG/ ← [Q-RAG](https://github.com/griver/Q-RAG.git)
└── datasets/ ← [datasets Hotpotqa and Musique](https://huggingface.co/datasets/Q-RAG/Hotpotqa_and_Musique)1git clone https://huggingface.co/datasets/Q-RAG/Hotpotqa_and_Musique
2cd Hotpotqa_and_Musique
3unzip hotpotqa+musique.zip -d /workspace/datasets
4cd ..
5rm -rf Hotpotqa_and_Musique
6du -h1git clone https://github.com/griver/Q-RAG.git
2cd Q-RAG
3#Only need when you don't have your self-trained hotpotqa model yet
4git clone https://huggingface.co/Q-RAG/qrag-ft-e5-on-hotpotqa1# Setup venv
2conda create -n qrag python=3.12 -y
3conda activate qrag
4
5python -m pip install -U pip wheel
6pip install vllm # pulls compatible PyTorch, Transformers, Triton, etc.
7pip install hydra-core tensorboard rotary-embedding-torch pandas nltk sortedcontainers accelerate datasets
8
9# Check environment
10python -c "from rl.agents.pqn import PQNActor; print('✅ Q-RAG installed successfully')"
111python train_q_rag_logt.py \
2 envs=hotpotqa \
3 algo=pqn_e5_hotpotqa \
4 envs.data_path="/workspace/datasets/hotpotqa" \
5 steps_count=10000 \
6 batch_size=12 \
7 accumulate_grads=8 \
8 eval_interval=50 \ #original 100
9 envs_parallel=1 \
10 max_action_length=2201python train_q_rag.py \
2 envs=hotpotqa \
3 algo=pqn_e5_hotpotqa \
4 envs.data_path="/workspace/datasets/hotpotqa" \
5 steps_count=10000 \
6 batch_size=12 \
7 accumulate_grads=8 \
8 eval_interval=100\
9 envs_parallel=1 \
10 max_action_length=220
11
