-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
101 lines (97 loc) · 4.44 KB
/
index.html
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
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Quickblink Projects</title>
<link rel="stylesheet" href="style.css">
<meta id="testViewport" name="viewport" content="width=device-width">
</head>
<body>
<div class="outercon">
<div id="titlecontainer">
<div class="title_fg desktop">My Projects</div>
<div class="title_fg mobile">Projects</div>
<!-- <div class="title_bg desktop">88:888888</div>
<div class="title_bg mobile">88:88</div> -->
<a href="https://github.com/quickblink">... to my Github profile</a>
</div>
<div class="projectcon">
<img class="projectimg" src="imgs/cascade_sample1.jpg"/>
<div class="projtitle">Cascade</div>
<div>
<ul>
<li>A mockup of a visual machine learning frame work</li>
<li>Visually arrange a computational graph in the web-based frontend and perform computations on the backend (similar to Jupyter)</li>
<li>A demo version lets you try it out without the backend - using pyodide to emulate python in javascript</li>
<li>Arbitrary python code can be executed</li>
<li>Many specific features are possible: Dynamic resource allocation, a history of training runs, smart hyperparameter management and more...
</li>
</ul>
</div>
<div class="flex">
<a class="github" href="https://github.com/Quickblink/cascade">
<div>
<img src="imgs/GitHub-Mark-Light-32px.png"/>
<p class="buttontext">Repository</p>
</div>
</a>
<a class="demo" href="cascade">Demo</a>
</div>
</div>
<div class="projectcon" id="rnnbuilder">
<div class="projectimg">
<img src="imgs/rnnbuilder_graph_3.svg"/>
</div>
<div class="projtitle">rnnbuilder</div>
<div>
<ul>
<li>Build models as a fixed computational graph in PyTorch in a way inspired by Keras</li>
<li>No separate __init__ and forward definitions necessary. Short and comprehensible model definitions
for complex and/or recurrent architectures
</li>
<li>Input sizes for each module as computed automatically based on the input shape to the network</li>
<li>All hidden state of submodules in managed automatically</li>
<li>Executes sequences of data in a batched fashion whenever possible and performs graph analysis to do
so
</li>
</ul>
<div class="pipins">
<span id="instruction" class="pipinner">pip install rnnbuilder</span>
<button onclick="myFunction()" id="copybttn" class="pipinner"><img src="imgs/copy-solid.svg"></button>
</div>
</div>
<div class="flex">
<a class="github" href="https://github.com/Quickblink/rnnbuilder">
<div>
<img src="imgs/GitHub-Mark-Light-32px.png"/>
<p class="buttontext">Repository</p>
</div>
</a>
<a class="demo" href="rnnbuilder">Documentation</a>
</div>
</div>
<div class="projectcon" id="ataricon">
<img class="projectimg" src="imgs/atari_sample2.png"/>
<div class="projtitle">RL framework</div>
<div>
<ul>
<li>A mostly general Q-learning framework</li>
<li>Multi-processing is used to parallelize data generation and training</li>
<li>Different environments can be used in a format close to gym</li>
<li>Recurrent networks are supported and their state is saved and dynamically updated in the replay buffer</li>
<li>Data can be generated according to multiple actor policies (greedy, deterministic, etc.) and combinations are possible to formulate "training schedules"</li>
<li>Frame stacking is supported</li>
<li>Other modifications to standard Q-learning are possible</li>
</ul>
</div>
<a class="github" href="https://github.com/Quickblink/RLframework">
<div>
<img src="imgs/GitHub-Mark-Light-32px.png"/>
<p class="buttontext">Repository</p>
</div>
</a>
</div>
</div>
<script src="script.js"></script>
</body>
</html>