Issues

• Trajectories.log_rewards lacks a detach guard: a differentiable state would silently change the TB estimator

  GFNOrg/torchgfn#527 · Filter by author josephdviviano opened on Aug 7, 2026

• implement this spinfoam environment

  GFNOrg/torchgfn#429 · Filter by author josephdviviano opened on Nov 3, 2025
  2

  

• Extend modules to support more kinds of diffusion processes

  GFNOrg/torchgfn#422 · Filter by author josephdviviano opened on Oct 23, 2025
  2

• Possible variants of the DiffusionSampling environment to add

  GFNOrg/torchgfn#421 · Filter by author josephdviviano opened on Oct 23, 2025

• Add policy gradient method

  enhancementhigh priority
  GFNOrg/torchgfn#416 · Filter by author josephdviviano opened on Oct 14, 2025
  

• Explore puttingget_trajectory/transition_* back into the estimator via a PolicyMixin

  GFNOrg/torchgfn#414 · Filter by author josephdviviano opened on Oct 14, 2025

• Some sources of inefficiency.

  GFNOrg/torchgfn#377 · Filter by author hyeok9855 opened on Aug 19, 2025
  

• Thompson sampling for improved exploration in GFlowNets

  enhancement
  GFNOrg/torchgfn#374 · Filter by author josephdviviano opened on Aug 10, 2025
  

• Improving Gradient-guided Nested Sampling for Posterior Inference

  enhancement
  GFNOrg/torchgfn#373 · Filter by author josephdviviano opened on Aug 10, 2025
  

• pass off policy hyperparameters as a vector with the length of a batch instead of a scalar

  GFNOrg/torchgfn#369 · Filter by author josephdviviano opened on Aug 1, 2025
  

• Experiment with different precision levels for performance?

  GFNOrg/torchgfn#366 · Filter by author josephdviviano opened on Jul 30, 2025

• Need a way of calculating posterior recovery in very high dimensional space -- i.e., by pre-computing all states once and saving these results to disk.

  GFNOrg/torchgfn#365 · Filter by author josephdviviano opened on Jul 30, 2025

• Method for sampling different off-policy exploration techniques (i.e., strategies for each policy to adopt)

  GFNOrg/torchgfn#361 · Filter by author josephdviviano opened on Jul 30, 2025
  1

  

• make edge/node features in graphs configurable as either float or int

  enhancement
  GFNOrg/torchgfn#356 · Filter by author josephdviviano opened on Jul 26, 2025
  

• List of potential applications of distributed torchgfn [TO COMPLETE]

  GFNOrg/torchgfn#344 · Filter by author josephdviviano opened on Jul 4, 2025
  

• Break set_seed() into multiple functions (for distributed and local cases respectively)

  GFNOrg/torchgfn#342 · Filter by author josephdviviano opened on Jul 3, 2025

• More tests to ensure no state mutation on a per-environment basis.

  GFNOrg/torchgfn#330 · Filter by author josephdviviano opened on Jun 25, 2025

• Sampling with backward policy recovers log_rewards of initial_states

  GFNOrg/torchgfn#315 · Filter by author alexandrelarouche opened on May 14, 2025
  6

• Proposal: Restructure README.md for Conciseness and Clarity

  high priority
  GFNOrg/torchgfn#309 · Filter by author saleml opened on Apr 19, 2025
  1

• Add a few notebooks on graphs, applications, and environment definition

  documentationenhancementhelp wanted
  GFNOrg/torchgfn#283 · Filter by author josephdviviano opened on Mar 18, 2025
  

• flow matching should allow for intermediate state reward calculation

  GFNOrg/torchgfn#242 · Filter by author josephdviviano opened on Feb 11, 2025
  

• flow matching tutorial notebook

  GFNOrg/torchgfn#229 · Filter by author saleml opened on Jan 19, 2025
  

• Add support for main features of GFN-Diffusion

  enhancementhelp wanted
  GFNOrg/torchgfn#218 · Filter by author josephdviviano opened on Nov 15, 2024
  

• Estimators make too many assumptions about the shapes of outputs.

  bughelp wantedhigh prioritysmall
  GFNOrg/torchgfn#214 · Filter by author josephdviviano opened on Nov 13, 2024
  1

  

• Issues in ReplayBuffer

  enhancementhelp wantedhigh priority
  GFNOrg/torchgfn#201 · Filter by author hyeok9855 opened on Oct 18, 2024
  12