Commits

Branch selector

 master

User selector

sgirgin

Datepicker

All time

Commit History

Commits on Apr 7, 2022

• Added a class for parametric distributions, defined a tabular distribution and

  Show description for 6e36388
  

  sgirgin

  authored and

  lanctot

  committed
  6e36388

  Copy full SHA for 6e36388

Commits on Mar 17, 2022

• Added gradient clipping to DQN agent for more stable training.

  Show description for 00f2f67
  

  sgirgin

  authored and

  lanctot

  committed
  00f2f67

  Copy full SHA for 00f2f67

Commits on Mar 15, 2022

• Added the example for deep fictitious play using DQN agents. This also changes

  Show description for 191c0f8
  

  sgirgin

  authored and

  lanctot

  committed
  191c0f8

  Copy full SHA for 191c0f8

Commits on Feb 27, 2022

• Updated the predator and prey game such that the string representations of the states are cached and the distribution is not (deep) copied when we clone a state (e.g. a child state is generated by …

  Show description for 2b7d32d
  

  sgirgin

  authored and

  lanctot

  committed
  2b7d32d

  Copy full SHA for 2b7d32d

• Optimizations for the dynamic routing game. In particular, we cache the string

  Show description for 0709cfc
  

  sgirgin

  authored and

  lanctot

  committed
  0709cfc

  Copy full SHA for 0709cfc

• Fixed log term when there are illegal actions in JAX implementation of DQN

  Show description for 809098f
  

  sgirgin

  authored and

  lanctot

  committed
  809098f

  Copy full SHA for 809098f

Commits on Feb 21, 2022

• Added separate example programs for running fictitious play and mirror descent

  Show description for 1848c4e
  

  sgirgin

  authored and

  lanctot

  committed
  1848c4e

  Copy full SHA for 1848c4e

• Added a factory method to create MFG games with pre-defined settings.

  Show description for cacb272
  

  sgirgin

  authored and

  lanctot

  committed
  cacb272

  Copy full SHA for cacb272

• Moved the method to create the default metric writer to a separate file under

  Show description for 76aad12
  

  sgirgin

  authored and

  lanctot

  committed
  76aad12

  Copy full SHA for 76aad12

• Added the implementation of Boltzmann DQN agent. We use the existing DQN agent

  Show description for 4099270
  

  sgirgin

  authored and

  lanctot

  committed
  4099270

  Copy full SHA for 4099270

• Added a utility function to generate the string representation of the forbidden

  Show description for 6ea9710
  

  sgirgin

  authored and

  lanctot

  committed
  6ea9710

  Copy full SHA for 6ea9710

Commits on Feb 13, 2022

• Implementation of the deep online mirror descent algorithm.

  Show description for 12ad213
  

  sgirgin

  authored and

  lanctot

  committed
  12ad213

  Copy full SHA for 12ad213

• Internal Google change

  Show description for 6e1447e
  

  sgirgin

  authored and

  lanctot

  committed
  6e1447e

  Copy full SHA for 6e1447e

• Ensure that player_id is converted to an integer when building the

  Show description for 7196ed8
  

  sgirgin

  authored and

  lanctot

  committed
  7196ed8

  Copy full SHA for 7196ed8

• Modified the crowd modelling game in 2D:

  Show description for 2945a26
  

  sgirgin

  authored and

  lanctot

  committed
  2945a26

  Copy full SHA for 2945a26

• The reservoir buffer class that was used in Neural Fictitious Self Play (NFSP)

  Show description for 58a98ab
  

  sgirgin

  authored and

  lanctot

  committed
  58a98ab

  Copy full SHA for 58a98ab

Commits on Jan 31, 2022

• Updated the observation tensor of the dynamic routing game state such that it

  Show description for 6d09e62
  

  sgirgin

  authored and

  lanctot

  committed
  6d09e62

  Copy full SHA for 6d09e62

Commits on Dec 13, 2021

• Updated the 2d crowd modelling environment such that forbidden states are taken

  Show description for eb801de
  

  sgirgin

  authored and

  lanctot

  committed
  eb801de

  Copy full SHA for eb801de

• Added a method to reset the contents of replay buffer.

  Show description for 42285ce
  

  sgirgin

  authored and

  lanctot

  committed
  42285ce

  Copy full SHA for 42285ce

Commits on Nov 23, 2021

• Moved ReplayBuffer under utils.

  Show description for 3b43f0b
  

  sgirgin

  authored and

  lanctot

  committed
  3b43f0b

  Copy full SHA for 3b43f0b

Commits on Nov 15, 2021

• Added methods get and update the distribution (for MFG games) in RL

  Show description for 5545991
  

  sgirgin

  authored and

  lanctot

  committed
  5545991

  Copy full SHA for 5545991

Commits on Nov 10, 2021

• Renamed DQNPolicies to JointRLAgentPolicy and moved to a separate file so that

  Show description for 5b149b2
  

  sgirgin

  authored and

  lanctot

  committed
  5b149b2

  Copy full SHA for 5b149b2

Commits on Nov 1, 2021

• Added a platform specific version of absl.app under python/utils.

  Show description for 9e91811
  

  sgirgin

  authored and

  lanctot

  committed
  9e91811

  Copy full SHA for 9e91811

Commits on Oct 27, 2021

• Updated the MFG DQN example to log the metrics using the clu library.

  Show description for 6ecb5f8
  

  sgirgin

  authored and

  lanctot

  committed
  6ecb5f8

  Copy full SHA for 6ecb5f8

Commits on Sep 27, 2021

• Update the MFG algorithms to use value functions. This will allow us to use

  Show description for 63cf573
  

  sgirgin

  authored and

  lanctot

  committed
  63cf573

  Copy full SHA for 63cf573