| DQN model applied to the this discrete environments CartPole-v1 | |
| ## Model Description | |
| The model was trained from the CleanRl library using the DQN algorithm | |
| ## Intended Use & Limitation | |
| The model is intended to be used for the following environments CartPole-v1 | |
| and understand the implication of Quantization on this type of model from a pretrained state## Training Procdure | |
| ### Training Hyperparameters | |
| ``` | |
| The folloing hyperparameters were used during training: | |
| - exp_name: functional_dqn | |
| - seed: 0 | |
| - torch_deterministic: True | |
| - cuda: False | |
| - track: True | |
| - wandb_project_name: cleanRL | |
| - wandb_entity: compress_rl | |
| - capture_video: False | |
| - env_id: CartPole-v1 | |
| - total_timesteps: 500000 | |
| - learning_rate: 0.00025 | |
| - buffer_size: 10000 | |
| - gamma: 0.99 | |
| - target_network_frequency: 500 | |
| - batch_size: 128 | |
| - start_e: 1 | |
| - end_e: 0.05 | |
| - exploration_fraction: 0.5 | |
| - learning_starts: 10000 | |
| - train_frequency: 10 | |
| - optimizer: Adam | |
| - wandb_project: cleanrl | |
| ``` | |
| ### Framework and version | |
| ``` | |
| Pytorch 1.12.1+cu102 | |
| gym 0.23.1 | |
| Weights and Biases 0.13.3 | |
| Hugging Face Hub 0.11.1 | |