|
--- |
|
library_name: diambra |
|
tags: |
|
- street-fighter-iii |
|
- deep-reinforcement-learning |
|
- reinforcement-learning |
|
- stable-baseline3 |
|
--- |
|
|
|
# Model Card for Model ID |
|
|
|
<!-- Provide a quick summary of what the model is/does. --> |
|
|
|
A DRL agent playing Street Fighter III trained using diambra ai library |
|
|
|
## Codes |
|
|
|
Github repos(Give a star if found useful): |
|
* https://github.com/hishamcse/Advanced-DRL-Renegades-Game-Bots |
|
* https://github.com/hishamcse/DRL-Renegades-Game-Bots |
|
* https://github.com/hishamcse/Robo-Chess |
|
|
|
|
|
## Model Details |
|
|
|
<!-- Provide the basic links for the model. --> |
|
|
|
- **My Code for this model:** https://github.com/hishamcse/Advanced-DRL-Renegades-Game-Bots/tree/main/VI%20-%20Diambra_AI_Street-Fighter-III |
|
- **Tutorial:** https://github.com/alexpalms/deep-rl-class/blob/main/units/en/unitbonus3 |
|
- **Documentation:** https://docs.diambra.ai/ |
|
|
|
## Training Details |
|
|
|
#### Training Hyperparameters |
|
|
|
``` |
|
folders: |
|
parent_dir: "./results/" |
|
model_name: "sr6_128x4_das_nc" |
|
|
|
settings: |
|
game_id: "sfiii3n" |
|
step_ratio: 6 |
|
frame_shape: !!python/tuple [128, 128, 1] |
|
continue_game: 0.0 |
|
action_space: "discrete" |
|
characters: "Ken" |
|
difficulty: 6 |
|
outfits: 2 |
|
|
|
wrappers_settings: |
|
normalize_reward: true |
|
no_attack_buttons_combinations: true |
|
stack_frames: 4 |
|
dilation: 1 |
|
add_last_action: true |
|
stack_actions: 12 |
|
scale: true |
|
exclude_image_scaling: true |
|
role_relative: true |
|
flatten: true |
|
filter_keys: ["action", "own_health", "opp_health", "own_side", "opp_side", "opp_character", "stage", "timer"] |
|
|
|
policy_kwargs: |
|
#net_arch: [{ pi: [64, 64], vf: [32, 32] }] |
|
net_arch: [64, 64] |
|
|
|
ppo_settings: |
|
gamma: 0.94 |
|
model_checkpoint: "0" # 0: No checkpoint, 100000: Load checkpoint (if previously trained for 100000 steps) |
|
learning_rate: [2.5e-4, 2.5e-6] # To start |
|
clip_range: [0.15, 0.025] # To start |
|
#learning_rate: [5.0e-5, 2.5e-6] # Fine Tuning |
|
#clip_range: [0.075, 0.025] # Fine Tuning |
|
batch_size: 512 #8 #nminibatches gave different batch size depending on the number of environments: batch_size = (n_steps * n_envs) // nminibatches |
|
n_epochs: 4 |
|
n_steps: 512 |
|
autosave_freq: 10000 |
|
time_steps: 100000 |
|
``` |
