Skip to content
Snippets Groups Projects
Select Git revision
  • 1e9062bbab4a93bd0fcd1053545434438ac768c9
  • master default protected
  • johannes
  • partial_parser
  • Aloui_Dary
  • ignore_punct
  • AC
  • classifier
  • fixhelp
  • libmacaon2
  • error_predictor
  • morpho
  • ssrnn
  • tfparsing
  • silvio
  • tagger_options
  • maca_trans_frame_parser
  • alexis
  • new_config
  • tagparse
  • maca_graph_parser
21 results

simple_decoder_parser_arc_eager.c

Blame
  • Rl.py 7.15 KiB
    import sys
    import random
    import torch
    import torch.nn.functional as F
    import numpy as np
    from Util import getDevice
    
    ################################################################################
    class ReplayMemory() :
      def __init__(self, capacity, stateSize) :
        self.capacity = capacity
        self.states = torch.zeros(capacity, stateSize, dtype=torch.long, device=getDevice())
        self.newStates = torch.zeros(capacity, stateSize, dtype=torch.long, device=getDevice())
        self.actions = torch.zeros(capacity, 1, dtype=torch.long, device=getDevice())
        self.rewards = torch.zeros(capacity, 1, device=getDevice())
        self.noNewStates = torch.zeros(capacity, dtype=torch.bool, device=getDevice())
        self.position = 0
        self.nbPushed = 0
    
      def push(self, state, action, newState, reward) :
        self.states[self.position] = state
        self.actions[self.position] = action
        if newState is not None :
          self.newStates[self.position] = newState
        self.noNewStates[self.position] = newState is None
        self.rewards[self.position] = reward 
        self.position = (self.position + 1) % self.capacity
        self.nbPushed += 1
    
      def sample(self, batchSize) :
        start = random.randint(0, len(self)-batchSize)
        end = start+batchSize
        return self.states[start:end], self.actions[start:end], self.newStates[start:end], self.noNewStates[start:end], self.rewards[start:end]
    
      def __len__(self):
        return min(self.nbPushed, self.capacity)
    ################################################################################
    
    ################################################################################
    def selectAction(network, state, ts, config, missingLinks, probaRandom, probaOracle) :
      sample = random.random()
      if sample < probaRandom :
        return ts[random.randrange(len(ts))]
      elif sample < probaRandom+probaOracle :
        candidates = sorted([[trans.getOracleScore(config, missingLinks), trans] for trans in ts if trans.appliable(config)])
        return candidates[0][1] if len(candidates) > 0 else None
      else :
        with torch.no_grad() :
          output = network(torch.stack([state]))
          predIndex = int(torch.argmax(output))
          return ts[predIndex]
    ################################################################################
    
    ################################################################################
    def optimizeModel(batchSize, policy_net, target_net, memory, optimizer, gamma) :
      if len(memory) < batchSize :
        return 0.0
    
      states, actions, nextStates, noNextStates, rewards = memory.sample(batchSize)
    
      predictedQ = policy_net(states).gather(1, actions)
      nextQ = target_net(nextStates).max(1)[0].detach().unsqueeze(0)
      nextQ = torch.transpose(nextQ, 0, 1)
      nextQ[noNextStates] = 0.0
    
      expectedReward = gamma*nextQ + rewards
    
      loss = F.smooth_l1_loss(predictedQ, expectedReward)
      optimizer.zero_grad()
      loss.backward()
      for param in policy_net.parameters() :
        if param.grad is not None :
          param.grad.data.clamp_(-1, 1)
      optimizer.step()
    
      return float(loss)
    ################################################################################
    
    ################################################################################
    def rewarding(appliable, config, action, missingLinks, funcname):
      return globals()["reward"+funcname](appliable, config, action, missingLinks)
    ################################################################################
    
    ################################################################################
    def rewardA(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -1.0*action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          error_in_pop = [i for i in range(1,back) if config.historyPop[-i][3] < 0]
          last_error = error_in_pop[-1] if len(error_in_pop) > 0 else 0
          reward = last_error - back
      else:
        reward = -3.0
      return reward
    ################################################################################
    
    ################################################################################
    def rewardB(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = 1.0 - action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          error_in_pop = [i for i in range(1,back) if config.historyPop[-i][3] < 0]
          last_error = error_in_pop[-1] if len(error_in_pop) > 0 else 0
          reward = last_error - back
      else:
        reward = -3.0
      return reward
    ################################################################################
    
    ################################################################################
    def rewardC(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          error_in_pop = [i for i in range(1,back) if config.historyPop[-i][3] < 0]
          canceledRewards = [h[3] for h in config.historyPop[-back:]]
          reward = -sum(canceledRewards)
      else:
        reward = -3.0
      return reward
    ################################################################################
    
    ################################################################################
    def rewardD(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          error_in_pop = [i for i in range(1,back) if config.historyPop[-i][3] < 0]
          canceledRewards = [h[3] for h in config.historyPop[-back:]]
          reward = -sum(canceledRewards) - 1
      else:
        reward = -3.0
      return reward
    ################################################################################
    
    ################################################################################
    def rewardE(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -action.getOracleScore(config, missingLinks)
        else :
          reward = -0.5
      else:
        reward = -3.0
      return reward
    ################################################################################
    
    ################################################################################
    def rewardF(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -1.0*action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          error_in_pop = [i for i in range(1,back) if config.historyPop[-i][3] < 0]
          last_error = error_in_pop[-1] if len(error_in_pop) > 0 else 0
          reward = last_error - back
      else:
        reward = -3.0
      return 10*reward
    ################################################################################
    
    ################################################################################
    def rewardG(appliable, config, action, missingLinks):
      if appliable:
        if "BACK" not in action.name :
          reward = -action.getOracleScore(config, missingLinks)
        else :
          back = action.size
          canceledRewards = [h[3] for h in config.historyPop[-back:]]
          reward = np.log(1-sum(canceledRewards)) if -sum(canceledRewards) > 0 else -1
      else:
        reward = -3.0
      return reward
    ################################################################################