Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
DQN model applied to the this discrete environments CartPole-v1
|
| 2 |
+
## Model Description
|
| 3 |
+
The model was trained from the CleanRl library using the DQN algorithm
|
| 4 |
+
## Intended Use & Limitation
|
| 5 |
+
The model is intended to be used for the following environments CartPole-v1
|
| 6 |
+
and understand the implication of Quantization on this type of model from a pretrained state## Training Procdure
|
| 7 |
+
### Training Hyperparameters
|
| 8 |
+
```
|
| 9 |
+
The folloing hyperparameters were used during training:
|
| 10 |
+
- exp_name: functional_dqn
|
| 11 |
+
- seed: 0
|
| 12 |
+
- torch_deterministic: True
|
| 13 |
+
- cuda: False
|
| 14 |
+
- track: True
|
| 15 |
+
- wandb_project_name: cleanRL
|
| 16 |
+
- wandb_entity: compress_rl
|
| 17 |
+
- capture_video: False
|
| 18 |
+
- env_id: CartPole-v1
|
| 19 |
+
- total_timesteps: 500000
|
| 20 |
+
- learning_rate: 0.00025
|
| 21 |
+
- buffer_size: 10000
|
| 22 |
+
- gamma: 0.99
|
| 23 |
+
- target_network_frequency: 500
|
| 24 |
+
- batch_size: 128
|
| 25 |
+
- start_e: 1
|
| 26 |
+
- end_e: 0.05
|
| 27 |
+
- exploration_fraction: 0.5
|
| 28 |
+
- learning_starts: 10000
|
| 29 |
+
- train_frequency: 10
|
| 30 |
+
- optimizer: Adam
|
| 31 |
+
- wandb_project: cleanrl
|
| 32 |
+
```
|
| 33 |
+
### Framework and version
|
| 34 |
+
```
|
| 35 |
+
Pytorch 1.12.1+cu102
|
| 36 |
+
gym 0.23.1
|
| 37 |
+
Weights and Biases 0.13.3
|
| 38 |
+
Hugging Face Hub 0.11.1
|