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>

1 files changed, 47 insertions, 0 deletions

diff --git a/collaborativeagents/training/grpo_verl/outputs/2026-01-11/03-50-42/.hydra/overrides.yaml b/collaborativeagents/training/grpo_verl/outputs/2026-01-11/03-50-42/.hydra/overrides.yaml
new file mode 100644
index 0000000..8c6f184
--- /dev/null
+++ b/collaborativeagents/training/grpo_verl/outputs/2026-01-11/03-50-42/.hydra/overrides.yaml
@@ -0,0 +1,47 @@
+- algorithm.adv_estimator=grpo
+- data.train_files=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training/grpo_verl/data/session_level_reflection_grpo_train.parquet
+- data.val_files=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training/grpo_verl/data/session_level_reflection_grpo_train.parquet
+- data.train_batch_size=64
+- data.max_prompt_length=2048
+- data.max_response_length=1024
+- data.filter_overlong_prompts=True
+- data.truncation=error
+- data.prompt_key=prompt
+- data.reward_fn_key=data_source
+- actor_rollout_ref.model.path=/work/nvme/bfqt/yurenh2/sft_checkpoints/checkpoint-200
+- actor_rollout_ref.actor.optim.lr=1e-6
+- actor_rollout_ref.model.use_remove_padding=True
+- actor_rollout_ref.actor.ppo_mini_batch_size=8
+- actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=4
+- actor_rollout_ref.actor.use_kl_loss=True
+- actor_rollout_ref.actor.kl_loss_coef=0.003
+- actor_rollout_ref.actor.kl_loss_type=low_var_kl
+- actor_rollout_ref.actor.entropy_coeff=0
+- actor_rollout_ref.model.enable_gradient_checkpointing=True
+- actor_rollout_ref.actor.fsdp_config.model_dtype=bfloat16
+- actor_rollout_ref.actor.fsdp_config.param_offload=False
+- actor_rollout_ref.actor.fsdp_config.optimizer_offload=False
+- actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=4
+- actor_rollout_ref.rollout.tensor_model_parallel_size=1
+- actor_rollout_ref.rollout.name=vllm
+- actor_rollout_ref.rollout.gpu_memory_utilization=0.5
+- actor_rollout_ref.rollout.n=8
+- actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=4
+- actor_rollout_ref.ref.fsdp_config.model_dtype=bfloat16
+- actor_rollout_ref.ref.fsdp_config.param_offload=True
+- actor_rollout_ref.rollout.temperature=0.9
+- actor_rollout_ref.rollout.top_p=0.9
+- custom_reward_function.path=/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training/grpo_verl/verl_reward_functions.py
+- custom_reward_function.name=compute_score
+- algorithm.use_kl_in_reward=False
+- trainer.critic_warmup=0
+- trainer.val_before_train=False
+- trainer.logger=["console"]
+- trainer.project_name=collaborative-agent-reflection-grpo
+- trainer.experiment_name=llama3.1-8b-grpo
+- trainer.n_gpus_per_node=2
+- trainer.nnodes=1
+- trainer.save_freq=50
+- trainer.test_freq=100
+- trainer.total_epochs=1
+- trainer.default_local_dir=/scratch/bfqt/yurenh2/grpo_outputs