kzl/decision-transformer

Issues

• Potential bug: Attention mask allows access to future tokens?

  #30 opened 2 years ago by donthomasitos
  3

• position embeddings do not vary between various time steps

  #64 opened a month ago by udaymallappa
  1

• MultiDiscrete Action Space

  #71 opened a month ago by hamzaali98
  1

• What should return to go be during inferencing?

  #70 opened a month ago by FinAminToastCrunch
  2

• Question about the output of the decision transformer

  #67 opened a month ago by Pulsar110
  3

• AttributeError: 'GPT2Config' object has no attribute 'n_ctx'

  #56 opened a year ago by YTL7
  6

• Where are the datasets?

  #74 opened 2 months ago by wilhem
  1

• compilation error (MuJoCo)

  #72 opened 2 months ago by NagarajJ111
  0

• Gsutil Error

  #60 opened a year ago by ChaselWan
  1

• TypeError on trajectory_gpt2.py

  #31 opened 2 years ago by dljzx
  2

• Error when loading fixed replay buffer

  #34 opened 2 years ago by dgjung0220
  11

• Is there a code for the shortest path search?

  #48 opened 2 years ago by geon0325
  1

• No registered env with id: halfcheetah-medium-v2

  #44 opened 2 years ago by boykac
  4

• Image Data

  #65 opened 9 months ago by AsadMir10
  0

• Any reason for using different models for gym and atari experiments?

  #47 opened 2 years ago by GilgameshD
  1

• how to print accurate in this code?

  #63 opened a year ago by YuZheng23
  2

• torch.stack bug

  #62 opened a year ago by ChaselWan
  0

• why sample from multinomial distribution during evaluation in Atari?

  #61 opened a year ago by sallyqiansun
  0

• Problem Creating Deterministic Action Selection

  #39 opened 2 years ago by carsonsmith87
  1

• Questions about dataset preprocessing

  #55 opened a year ago by typoverflow
  0

• [IDEA] Code for dataset generation

  #54 opened a year ago by JustinS6626
  1

• Bug in state and action prediction

  #53 opened a year ago by ezhang7423
  0

• Global position embedding and timesteps look wrong in atari

  #51 opened a year ago by nzw0301
  2

• Why the padding is different for state, action, reward?

  #50 opened 2 years ago by CeyaoZhang
  0

• Padding / attention_mask questions

  #36 opened 2 years ago by DaveyBiggers
  2

• The setting of final token

  #49 opened 2 years ago by IpadLi
  0

• Atari results

  #46 opened 2 years ago by TongZhangTHU
  4

• MemoryError: Unable to allocate 6.57 GiB for an array with shape (7056000000,) and data type uint8

  #45 opened 2 years ago by leeruibin
  1

• Can you add reacher dataset

  #29 opened 3 years ago by jsun57
  1

• I am wondering if the shortest path case is included in the code? thanks

  #43 opened 2 years ago by bradley-code-again
  0

• The results on Mujoco reported in paper might be heavily influenced by env version

  #42 opened 2 years ago by linprophet
  0

• Padding tokens represented differently in different parts of the code

  #41 opened 2 years ago by asmadotgh
  0

• More Training Information on Reacher

  #40 opened 2 years ago by ivo-1
  0

• Confusion over shape of returns_to_go in get_batch

  #38 opened 2 years ago by DaveyBiggers
  0

• Some problems after reading the paper and code

  #37 opened 2 years ago by lylwy
  0

• Where are the weights for the trained models?

  #35 opened 2 years ago by simoninithomas
  1

• About parameters in code

  #32 opened 2 years ago by lylwy
  1

• Minor bug that removes best performing trajectory in gym experiments

  #33 opened 2 years ago by micahcarroll
  1

• about graph experiment

  #28 opened 3 years ago by ysymyth
  0

• batch sampling: only last tokens?

  #27 opened 3 years ago by Howuhh
  1

• undestanding use of rewards

  #24 opened 3 years ago by jeweinb
  2

• State and Return preds input

  #25 opened 3 years ago by backpropper
  1

• Question: is it possible to use the same Decision Transformer for new training trajectories generation?

  #26 opened 3 years ago by danielgafni
  2

• Return-to-go conditioning on Atari

  #23 opened 3 years ago by geekyutao
  2

• aligning action embeddings to other embeddings at line 237

  #20 opened 3 years ago by loct824
  3

• Possible misalignment in calculating rtg in Atari

  #22 opened 3 years ago by geekyutao
  1

• Regarding atari breakout results

  #21 opened 3 years ago by dido1998
  1

• difference between two GPT models used in this repo?

  #17 opened 3 years ago by ChenDRAG
  1

• Timesteps Shape

  #19 opened 3 years ago by MrShininnnnn
  1

• how to get the score of an expert policy and some other details

  #16 opened 3 years ago by TianhongDai
  1