update docs

Author: yusx-swapp

.DS_Store

@@ -1,2 +1,4 @@
 # Project exclude paths
-/gnnrl/data/pytorch_cifar100/
+/gnnrl/data/pytorch_cifar100/
+/site
+.idea

.gitignore

@@ -65,77 +65,3 @@ Will release soon
 [comment]: <> (export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/work/LAS/jannesar-lab/yusx/anaconda3/lib/)
 
 [comment]: <> (## Install from pip)
-
-<!-- 
-Currently we support installation on Linux, Mac and Windows. We also allow you to use docker.
-
-DGL works with the following operating systems:
-
-Ubuntu 16.04
-
-macOS X
-
-Windows 10
-
-DGL requires Python version 3.6, 3.7, 3.8 or 3.9.
-
-DGL supports multiple tensor libraries as backends, e.g., PyTorch, MXNet. For requirements on backends and how to select one, see Working with different backends.
-
-Starting at version 0.3, DGL is separated into CPU and CUDA builds. The builds share the same Python package name. If you install DGL with a CUDA 9 build after you install the CPU build, then the CPU build is overwritten.
-
-
-
-## Install by Conda
-
-
-Lorem markdownum Pyrrhus piscosamque? Sis Caras clamant **duorum**, somnia
-comitant, solvit, quos secum. Dato ipse Pergama ulla, accedere abigitque
-**aper**. Haec socero quoque ibi, cremarat atque amo qui tetigere crine,
-temptanti.
-
-- A cunctisque amens
-- Sospes genuumque pereat
-- Corpusque nam obice nec suis receptus Athos
-- Quam accipe raptam
-
-Diuque ut Perseia fabula foedantem quondam. Non sub simulatoremque iuvenci.
-
-    if (enterpriseForumBitrate - fiDot <= simmModel) {
-        ctr(ajaxRupHttps, mouse(technologyPppoe, reciprocal_servlet),
-                services_image_adc);
-        rightGopherApple += hover_desktop_adc.processFriendly(-3, compression,
-                scrollBootProtector);
-        clone_integrated(-5);
-    } else {
-        olePci(ergonomics_sli, urlReal, coreCpaGamma);
-        barcraftReimageCursor = applet - 4 + samplingPrinter;
-        maximize(uat_gate - terminalUnfriendSip, hyperlinkDaemonIm(handle_drag,
-                third, lifo), prebindingClient);
-    }
-    var rosetta_secondary_affiliate = firewire;
-    rjBus = 331673;
-
-## Install from Pip
-
-Orbem minacia turba minanti [afflatibus](http://simuldicitur.org/quoque.html)
-tremulo **iussit**, ab arsit relinquunt, poplite superos. Amplexumque retenta
-radice per Ityn est positos vulnus; *quae* anima ab inpulsu minores; ferenda
-[nec](http://egonostra.io/satavidamque.aspx) tellure. Labor momentaque Herculeae
-potiunda a **neque sanguine**, non quod defensae. Non mandata vinclo?
-
-    home.dvdPayloadHard += property_uml(barDirectMinimize);
-    if (4 + control > switch(shell)) {
-        terminalItunes.biometricsRemoteServer = pipeline_speed_click;
-        view_vector += web;
-    }
-    menuRoomCut -= core_postscript_direct + processorPc + drop_kindle;
-
-Dedecus haec; hic et recuset mens totidem inque, ut et quid capillis, ego erit
-alendum. Evehor cum color, vulnere, ipsaque non ponto, est acrior inmitis quae
-tinxi, ab.
-
-## Install from Source Code
-Invitus uvis preces mihi? Quam etiam en utque; alit et pro induroque dare
-profecturas iuvenum sollertia reperta resedit et traiecit gaude. Et non nymphae
-**utero temerarius cernit** egimus omnipotens hos, terrae, captum sanguine per;
-hiatu. -->

docs/Installation.md

@@ -0,0 +1,3 @@
+## Supported packages
+GNN-RL support popular deep graph neural network package, such as [Torch-Geometric](https://pytorch-geometric.readthedocs.io/en/latest/) and [DGL](https://www.dgl.ai/).
+In this section we will give examples to modeling DNN's topology to computational graph, and embedding them using graph neural network.

docs/graph/pyg.md

@@ -11,28 +11,12 @@ We have successfully tested the GNN-RL on model compression task, where we model
 
 We have tested our GNN-RL on fine-grained pruning and structured pruning on CNNs.
 We first model the trage DNN as a computational graph, and GNN-RL to search pruning policy directly through DNN's topology. In the reinforcement learning task definition, we use the computational graph as environment states, the action spaces are defined as pruning ratios, and the rewards are the compressed DNN's accuracy. Once the compressed model size satisfied the resource requirements, GNN-RL end the search episode. The network pruning task can be visualize as the diagram below.
-<!-- Se validi [marmor](http://www.subibis.net/) non si quoque minuuntur tergo,
-revelli **tenebris**, apex *Tethys* rogarem temperius monte altaque cura. Gratia
-molliri tempore, tanto mugitibus ictus. Iunctum *requirere probat* destinat
-vigore?
 
-- Mea per dum ruent invita quos et
-- Dicentum nece
-- Sibi iuro omnia sentit in timeo brevissimus
-- Misit adflat suum inposito vocem illic figuris -->
 
 ## Apply GNN-RL Pipeline on Other Discipline
 The GNN-RL is not limited to model compression, it aims to extract topology information through reinforcement learning task.
 We provide APIs to help you define your customized job!  It only requires your research object is a graph or been modeled as a graph, you can define your customized RL task (e.g., action space, environment states, rewards). Currently, our collegues are testing GNN-RL on job scheduling taks.
-<!-- 
-Rogos indotata geminas gaudebat ferendo, nemus quod multum lumina invocat
-tempora nebulae, et agnoscis! Pudori vulnere. Celerem festinus: delere currum
-venerabile limina spatiantia vastum, concita, mei Aeacides, et dea nefas. Artis
-fuit ille nostri quater lumina nec pectora Ixiona confessasque nostra et!
 
-> Sociis potentem summo, tamen consistere *amplexa in* pendere rursus nivosos. A
-> herbas excitus et tamen ego manibus ferebat parte. Acta dedit, e occursu
-> ferula in nomina laesi: suos. Crura iacens ora, tum ter officium nasci. -->
 
 ## Page Index
 Here is the index for you to quick locate the GNN-RL:
@@ -41,24 +25,7 @@ Here is the index for you to quick locate the GNN-RL:
 2. [Introduction by example](intro.md)
 3. [Model DNN as graph](graph/example.md)
 4. [GNN-RL in model compression](compression/pruning.md)
-<!-- Avis gratia, est illa est inrita propiora suum **nunc** apte mulcebat et est.
-Pallados Iuppiter pererrant tu alios repetiti flexisque nec turbavere mutare
-adpositi nec illis vertice, illo Phinea mihi. Dentibus *nece*. Angues in sedit
-spemque lapillos [praecipue](http://novaet.org/tempora.php) ego hos vulnera
-dictis.
-
-    ram_language = rawMotion.homeCarrierSystem(express);
-    fiT = hard_token;
-    drop_layout_version.ipv = in + javaIcqArchitecture(camelcaseBootImage,
-            hot_file, gpu_text);
-    day_table.pointSmmDynamic = megapixelListserv;
-
-Face tentoria in **quippe nymphe** praesepibus solacia moverat adfusaque callida
-dominum. Mors equus his et dique, est decimo quaerit, **quas arte**, omnes
-patent sequitur, superamur mergeret nil. **Ne** licet contigit victus ad
-carpebam ducit, feras et cum quoque parente! Sibi sorori mihi aura cetera
-propago praesignis Melanthus **decidit** esse omnia dedere domitamque quis
-cornum superis. -->
+
 
 ## Related Papers
 If you find our useful or use our work in your job, please cite the paper: