Hugging Face's logo

hishamcse
/
street-fighter-iii-ppo-diambra

Reinforcement Learning
diambra
street-fighter-iii
deep-reinforcement-learning
stable-baseline3
Model card Files Community
street-fighter-iii-ppo-diambra
File size: 2,242 Bytes 
bdbdc39
 
 
 
b26639a
 
 
 
 
90adb80
 
 
 
 
 
40e4817
 
 
 
 
4e4ade0
40e4817
 
90adb80
 
 
 
5f856f6
 
 
90adb80
 
 
 
 
cf85dab
90adb80
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
cf85dab
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
 
---
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
```