All pull requests

Pull requests

• stop_for_resubmission_when_low_time_left, coordinated stop, SLURM-level interruption

  #1992·albertz opened Oct 12, 2026Member·

  ·

• RF packed, one helper for an operand in the packing of another

  #1989·albertz opened Oct 11, 2026Member·

  ·

• RF monotonic RNN-T loss, blank scale and label smoothing

  #1988·polycrit opened Oct 11, 2026Contributor·

  ·

• PT load_optimizer: keep the state of params moved between param groups

  #1984·polycrit opened Oct 10, 2026Contributor·

  ·

• PT slice_update in place

  #1982·polycrit opened Oct 10, 2026Contributor·

  ·

• RF monotonic RNN-T loss, lattice log probs

  #1981·polycrit opened Oct 10, 2026Contributor·

  ·

• RF packed shift keeps the packing

  #1979·polycrit opened Oct 10, 2026Contributor·

  ·

• PT distributed sync levels, hierarchical grad / param sync

  #1976·albertz opened Oct 9, 2026Member·

  ·

• CachedDataset2, epoch without seqs with unknown num seqs

  #1973·polycrit opened Oct 8, 2026Contributor·

  ·

• MultiProcDataset, dedicated sharding yields every seq

  #1972·polycrit opened Oct 8, 2026Contributor·

  ·

• PT graph capture bounds from the batching options

  #1971·polycrit opened Oct 8, 2026Contributor·

  ·

• RF packed ctc_loss, shared target length

  #1970·polycrit opened Oct 8, 2026Contributor·

  ·

• RF packed ctc_best_path

  #1969·polycrit opened Oct 8, 2026Contributor·

  ·

• RF packed gather, plain source along a per-sequence axis

  #1968·polycrit opened Oct 8, 2026Contributor·

  ·

• RF packed gather, indices without the gathered dim

  #1965·polycrit opened Oct 8, 2026Contributor·

  ·

• RF Loss inv norm factor, own wrapper in the cache

  #1963·polycrit opened Oct 8, 2026Contributor·

  ·

• PT graph capture, compile_fx call shim from torch 2.11

  #1962·polycrit opened Oct 7, 2026Contributor·

  ·

• RF RNN-T full-sum loss over the packed lattice

  #1961·polycrit opened Oct 7, 2026Contributor·

  ·

• RF packed: declared Dim.packed_total_bound caps static buffer sizes

  #1955·robin-p-schmitt opened Oct 7, 2026Contributor·

  ·

• PT Triton RNN-T full-sum loss

  #1941·polycrit opened Oct 6, 2026Contributor·

  ·

• PT engine: torch_preload_next_train_epoch, prepare next train epoch data during eval

  #1939·albertz opened Oct 6, 2026Member·

  ·
  1

• PostprocessingDataset, predefined seq order for a per-seq stream

  #1937·polycrit opened Oct 6, 2026Contributor·

  ·
  3

• RF packed attention keeps the query packing

  #1933·polycrit opened Oct 5, 2026Contributor·

  ·

• PT schedule-free optimizer train and eval hooks

  #1925·polycrit opened Oct 4, 2026Contributor·

  ·

• PT MultiOptimizer over param subsets

  #1924·polycrit opened Oct 4, 2026Contributor·

  ·
  1