Upload README.md with huggingface_hub
Browse files
README.md
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
DQN model applied to the this discrete environments CartPole-v1
|
5 |
+
## Model Description
|
6 |
+
The model was trained from the CleanRl library using the DQN algorithm on CartPole-v1
|
7 |
+
## Intended Use & Limitation
|
8 |
+
The model is intended to be used for the following environments CartPole-v1
|
9 |
+
and understand the implication of Quantization on this type of model from a pretrained state
|
10 |
+
## Training Procdure
|
11 |
+
### Training Hyperparameters
|
12 |
+
The folloing hyperparameters were used during training:
|
13 |
+
- exp_name: functional_dqn
|
14 |
+
- seed: 0
|
15 |
+
- torch_deterministic: True
|
16 |
+
- cuda: False
|
17 |
+
- track: True
|
18 |
+
- wandb_project_name: cleanRL
|
19 |
+
- wandb_entity: compress_rl
|
20 |
+
- capture_video: False
|
21 |
+
- env_id: CartPole-v1
|
22 |
+
- total_timesteps: 500000
|
23 |
+
- learning_rate: 0.00025
|
24 |
+
- buffer_size: 10000
|
25 |
+
- gamma: 0.99
|
26 |
+
- target_network_frequency: 500
|
27 |
+
- batch_size: 128
|
28 |
+
- start_e: 1
|
29 |
+
- end_e: 0.05
|
30 |
+
- exploration_fraction: 0.5
|
31 |
+
- learning_starts: 10000
|
32 |
+
- train_frequency: 10
|
33 |
+
- optimizer: Adan
|
34 |
+
- max_grad_norm: 0.0
|
35 |
+
- weight_decay: 0.02
|
36 |
+
- opt_eps: None
|
37 |
+
- opt_betas: None
|
38 |
+
- no_prox: False
|
39 |
+
- wandb_project: cleanrl
|
40 |
+
### Framework and version
|
41 |
+
Pytorch 1.12.1+cu102
|
42 |
+
|
43 |
+
gym 0.23.1
|
44 |
+
Weights and Biases 0.13.3
|
45 |
+
Hugging Face Hub 0.11.1
|
46 |
+
Python Version 3.8.16 (default, Dec 7 2022, 01:12:13)
|
47 |
+
[GCC 7.5.0]
|
48 |
+
## Citation
|
49 |
+
```bibtex
|
50 |
+
```
|