Merge pull request #16 from twitter/kmathewson/rename

refactored name of luarocks package from rl to twrl

CMakeLists.txt

@@ -5,7 +5,7 @@ FIND_PACKAGE(Torch REQUIRED)
 
 FILE(GLOB luasrc src/*.lua)
 
-ADD_TORCH_PACKAGE(rl "" "${luasrc}")
+ADD_TORCH_PACKAGE(twrl "" "${luasrc}")
 
-INSTALL(DIRECTORY "src/agent" DESTINATION "${Torch_INSTALL_LUA_PATH_SUBDIR}/rl")
-INSTALL(DIRECTORY "src/gym-http-api/binding-lua" DESTINATION "${Torch_INSTALL_LUA_PATH_SUBDIR}/rl")
+INSTALL(DIRECTORY "src/agent" DESTINATION "${Torch_INSTALL_LUA_PATH_SUBDIR}/twrl")
+INSTALL(DIRECTORY "src/gym-http-api/binding-lua" DESTINATION "${Torch_INSTALL_LUA_PATH_SUBDIR}/twrl")

examples/run.lua

@@ -62,4 +62,4 @@ local nSteps, nIterations = params.nSteps, params.nIterations
 params.outdir = logDir .. '/gym'
 
 -- run test
-local performance = require 'rl.experiment'(env, agent, nSteps, nIterations, params)
+local performance = require 'twrl.experiment'(env, agent, nSteps, nIterations, params)

src/agent/baseAgent.lua

@@ -15,7 +15,7 @@ local function getAgent(opt)
    opt.nHiddenLayerSize = opt.nHiddenLayerSize or 10
    if opt.model then
       local modelName = opt.model
-      model = require('rl.agent.model.' .. opt.model)({
+      model = require('twrl.agent.model.' .. opt.model)({
          nInputs = envDetails.nbStates,
          nOutputs = envDetails.nbActions,
          nHiddenLayerSize = opt.nHiddenLayerSize,
@@ -27,7 +27,7 @@ local function getAgent(opt)
       })
    end
 
-   policy = require('rl.agent.policy')[opt.policy]({
+   policy = require('twrl.agent.policy')[opt.policy]({
       client = opt.client,
       instanceID = instanceID,
       envDetails = envDetails,
@@ -41,7 +41,7 @@ local function getAgent(opt)
       randomActionSampler = randomActionSampler
    })
 
-   local learn = require('rl.agent.learningUpdate.' .. opt.learningUpdate)({
+   local learn = require('twrl.agent.learningUpdate.' .. opt.learningUpdate)({
       model = model,
       envDetails = envDetails,
       tdLearnUpdate = opt.tdLearnUpdate,

src/agent/init.lua

@@ -1,8 +1,8 @@
 local agent = {
-   baseAgent = require 'rl.agent.baseAgent',
-   learningUpdate = require 'rl.agent.learningUpdate',
-   model = require 'rl.agent.model',
-   policy = require 'rl.agent.policy'
+   baseAgent = require 'twrl.agent.baseAgent',
+   learningUpdate = require 'twrl.agent.learningUpdate',
+   model = require 'twrl.agent.model',
+   policy = require 'twrl.agent.policy'
 }
 
 return agent

src/agent/learningUpdate/init.lua

@@ -1,8 +1,8 @@
 local learningUpdate = {
-   noLearning = require 'rl.agent.learningUpdate.noLearning',
-   reinforce = require 'rl.agent.learningUpdate.reinforce',
-   tdLambda = require 'rl.agent.learningUpdate.tdLambda',
-   util = require 'rl.agent.learningUpdate.util'
+   noLearning = require 'twrl.agent.learningUpdate.noLearning',
+   reinforce = require 'twrl.agent.learningUpdate.reinforce',
+   tdLambda = require 'twrl.agent.learningUpdate.tdLambda',
+   util = require 'twrl.agent.learningUpdate.util'
 }
 
 return learningUpdate

src/agent/learningUpdate/reinforce.lua

@@ -9,7 +9,7 @@ local function getLearningUpdate(opt)
    local gradClip = opt.gradClip
    local optim = require 'optim'
    local mo = require 'moses'
-   local util = require 'rl.agent.learningUpdate.util'
+   local util = require 'twrl.agent.learningUpdate.util'
    local smallEps = util.smallEps
    local model = opt.model
    local net = model.net

src/agent/model/init.lua

@@ -1,8 +1,8 @@
 local model = {
-   mlp = require 'rl.agent.model.mlp',
-   noModel = require 'rl.agent.model.noModel',
-   qFunction = require 'rl.agent.model.qFunction',
-   tilecoding = require 'rl.agent.model.tilecoding'
+   mlp = require 'twrl.agent.model.mlp',
+   noModel = require 'twrl.agent.model.noModel',
+   qFunction = require 'twrl.agent.model.qFunction',
+   tilecoding = require 'twrl.agent.model.tilecoding'
 }
 
 return model

src/agent/model/qFunction.lua

@@ -24,7 +24,7 @@ local function getModel(opt)
    local stateScalingFactor, stateMins = getStateMinsAndScaling(envDetails, numTilings)
    local memorySize = numTiles * numTiles
 
-   local tc = require 'rl.agent.model.tilecoding'({numTilings = numTilings, memorySize = memorySize, scaleFactor = stateScalingFactor, stateMins = stateMins})
+   local tc = require 'twrl.agent.model.tilecoding'({numTilings = numTilings, memorySize = memorySize, scaleFactor = stateScalingFactor, stateMins = stateMins})
 
    local weights = torch.FloatTensor(numTilings * memorySize * nbActions):zero():fill(initialWeightVal)
    local eligibility = torch.FloatTensor(numTilings * memorySize * nbActions):zero():fill(0)

src/agent/policy/init.lua

@@ -1,14 +1,14 @@
-local actionSamplers = require 'rl.agent.policy.actionSamplers'
+local actionSamplers = require 'twrl.agent.policy.actionSamplers'
 
 local policy = {
-    egreedy = require 'rl.agent.policy.egreedy',
-    normal = require 'rl.agent.policy.stochasticModelPolicy'({
+    egreedy = require 'twrl.agent.policy.egreedy',
+    normal = require 'twrl.agent.policy.stochasticModelPolicy'({
              actionSampler = actionSamplers.normal
         }),
-    categorical = require 'rl.agent.policy.stochasticModelPolicy'({
+    categorical = require 'twrl.agent.policy.stochasticModelPolicy'({
              actionSampler = actionSamplers.categorical
         }),
-    random = require 'rl.agent.policy.random',
+    random = require 'twrl.agent.policy.random',
 }
 
 return policy

src/experiment.lua

@@ -1,6 +1,6 @@
 local function experiment(envName, agent, nSteps, nIterations, opt)
-   local util = require 'rl.util'()
-   local gymClient = require 'rl.binding-lua.gym_http_client'
+   local util = require 'twrl.util'()
+   local gymClient = require 'twrl.binding-lua.gym_http_client'
    local opt = opt or {}
    local base = 'http://127.0.0.1:5000'
    local client = gymClient.new(base)
@@ -11,7 +11,7 @@ local function experiment(envName, agent, nSteps, nIterations, opt)
    local resume = opt.resume
    local renderAllSteps = opt.renderAllSteps
    local render = renderAllSteps == 'true' and true or false
-   local perf = require 'rl.perf'({nIterations = nSteps, windowSize = opt.windowSize})
+   local perf = require 'twrl.perf'({nIterations = nSteps, windowSize = opt.windowSize})
 
    local function run()
       client:env_monitor_start(instanceID, outdir, force, resume, video)
@@ -24,7 +24,7 @@ local function experiment(envName, agent, nSteps, nIterations, opt)
          agentOpt.learningUpdate = agent.learningUpdate
          agentOpt.envDetails = util.getStateAndActionSpecs(agentOpt.stateSpace, agentOpt.actionSpace)
          function agentOpt.randomActionSampler() return client:env_action_space_sample(instanceID) end
-      local agent = require 'rl.agent.baseAgent'(agentOpt)
+      local agent = require 'twrl.agent.baseAgent'(agentOpt)
       local iterPerformance = {}
 
       for nIter = 1, nIterations do
@@ -50,7 +50,7 @@ local function experiment(envName, agent, nSteps, nIterations, opt)
       client:env_monitor_close(instanceID)
 
       if opt.uploadResults == 'true' then
-         print('Uploading results, check server for URL.')
+         print('Uploading results, check server for URL')
          -- Upload to the scoreboard, OPENAI_GYM_API_KEY must be set
          client:upload(outdir)
       end

src/init.lua

@@ -1,15 +1,15 @@
-local rl = {}
+local twrl = {}
 
 -- Meta info
-rl.VERSION = '0.1'
-rl.LICENSE = 'MIT'
+twrl.VERSION = '0.1'
+twrl.LICENSE = 'MIT'
 
 -- Utility packages 
-rl.agent = require 'rl.agent'
-rl.util = require 'rl.util'
-rl.perf = require 'rl.perf'
-rl.experiment = require 'rl.experiment'
-rl.gymClient = require 'rl.binding-lua.gym_http_client'
+twrl.agent = require 'twrl.agent'
+twrl.util = require 'twrl.util'
+twrl.perf = require 'twrl.perf'
+twrl.experiment = require 'twrl.experiment'
+twrl.gymClient = require 'twrl.binding-lua.gym_http_client'
 
 -- Return package
-return rl
+return twrl

test/test-gym.lua

@@ -46,7 +46,7 @@ function mujoco.testMujoco()
 end
 
 function experiment.badExperimentCall()
-   local performance = require 'rl.experiment'()
+   local performance = require 'twrl.experiment'()
    tester:eq(performance, {}, "bad experiment call should fail with bad settings ")
 end
 
@@ -60,7 +60,7 @@ function experiment.randomNoLearningNoModel()
    }
    local nSteps = 2
    local nIterations = 2
-   local performance = require 'rl.experiment'(env, agent, nSteps, nIterations, params)
+   local performance = require 'twrl.experiment'(env, agent, nSteps, nIterations, params)
    tester:eq(performance.iteration, 2, "basic experiment should run")
 end
 

test/test.lua

@@ -6,7 +6,7 @@ local tilecoding = torch.TestSuite()
 local tester = torch.Tester()
 
 -- Load all 
-local util = require 'rl.util'()
+local util = require 'twrl.util'()
 
 function base.torchTensor()
    local a = {2, torch.Tensor{1, 2, 2}}
@@ -19,7 +19,7 @@ function tilecoding.tilecodeConsistent()
    local numTiles = 32
    local memorySize = numTiles * numTiles
    local stateScalingFactor = {1, 1}
-   local tc = require 'rl.agent.model.tilecoding'(({
+   local tc = require 'twrl.agent.model.tilecoding'(({
       numTilings = numTilings, 
       memorySize = memorySize, 
       scaleFactor = stateScalingFactor
@@ -35,7 +35,7 @@ function tilecoding.tilecodePredictable()
    local numTiles = 32
    local memorySize = numTiles * numTiles
    local stateScalingFactor = {1, 1}
-   local tc = require 'rl.agent.model.tilecoding'(({
+   local tc = require 'twrl.agent.model.tilecoding'(({
       numTilings = numTilings, 
       memorySize = memorySize, 
       scaleFactor = stateScalingFactor
@@ -47,25 +47,25 @@ function tilecoding.tilecodePredictable()
 end
 
 function performance.reset()
-   local perf = require 'rl.perf'()
+   local perf = require 'twrl.perf'()
    local emptyTable = perf.reset()
    tester:eq(emptyTable, {}, "performance: reset failed")
 end
 
 function performance.addRewardTerminal()
-   local perf = require 'rl.perf'()
+   local perf = require 'twrl.perf'()
    local traj, trajs = perf.addReward(1, 1, true)
    tester:eq(traj, {}, "performance: add reward terminal failed")
 end
 
 function performance.addRewardNonTerminal()
-   local perf = require 'rl.perf'()
+   local perf = require 'twrl.perf'()
    local traj, trajs = perf.addReward(1, 1, false)
    tester:eq(traj, {1}, "performance: add reward non-terminal failed")
 end
 
 function performance.getSummary()
-   local perf = require 'rl.perf'({windowSize = 10})
+   local perf = require 'twrl.perf'({windowSize = 10})
    local _, _ = perf.addReward(1, 0, false)
    local _, _ = perf.addReward(1, 0, false)
    local _, _ = perf.addReward(1, 1, true)

rl-scm-1.rockspec → twrl-scm-1.rockspec

@@ -1,15 +1,15 @@
-package = "rl"
+package = "twrl"
 version = "scm-1"
 source = {
-   url = "https://github.com/twitter/torch-rl.git"
+   url = "https://github.com/twitter/torch-twrl.git"
 }
 description = {
    summary = "Reinforcement Learning for Torch and Lua",
    detailed = [[
-      torch-rl is a Reinforcement Learning framework
+      torch-twrl is a Reinforcement Learning framework
       built for Torch. It interfaces with OpenAI Gym.
    ]],
-   homepage = "https://github.com/twitter/torch-rl",
+   homepage = "https://github.com/twitter/torch-twrl",
    license = "MIT"
 }
 dependencies = {