fix: implement lazy transpose to resolve slow Tensor::transpose#436

Closed

Aakash-Naidu wants to merge 2339 commits intopaiml:mainfrom

Aakash-Naidu:fix/transpose-performance-388

This pull request is big! We're only showing the most recent 250 commits

Commits on Feb 28, 2026

fix: tensor evidence overrides metadata architecture claims (Refs GH-279)

noahgift
and
claude
committed

Commits on Mar 1, 2026

fix: add realizar path patch for BF16 embedding fix (Refs GH-187)

noahgift
and
claude
committed
fix: align BF16 dtype byte to GGML spec (BF16=30, not 2) (Refs GH-187)

noahgift
and
claude
committed
fix: distinguish HTTP 404 from other network errors in download_file() (Refs GH-356)

noahgift
and
claude
committed
docs: APR checkpoint spec v1.2.0 — full command matrix, 4 falsification rounds (Refs PMAT-342)

noahgift
and
claude
committed
feat: add AprReader::open_filtered() for inference tensor loading (F-CKPT-016)

noahgift
and
claude
committed
fix: add missing scores field to GgufTokenizer constructor (GH-366)

noahgift
and
claude
committed
docs: APR checkpoint spec v1.3.0 — mark 11/18 contracts implemented

noahgift
and
claude
committed
feat: atomic writes, NaN scan, shape validation in APR reader/writer

noahgift
and
claude
committed
test: add falsification tests for all APR checkpoint contracts

noahgift
and
claude
committed
feat: add SentencePiece tokenizer support for APR embedding (Refs GH-366)

noahgift
and
claude
committed
docs: update APR checkpoint spec to v1.4.0 (18/18 contracts done)

noahgift
and
claude
committed
fix: Load embedded BPE tokenizer in apr serve for GGUF-imported models (Refs GGUF-SERVE-001)

noahgift
and
claude
committed
feat: add `apr tokenize plan/apply` subcommand for BPE vocabulary training

noahgift
and
claude
committed
fix: Respect --quantize flag for F16/BF16 SafeTensors sources (Refs GH-205)

noahgift
and
claude
committed
feat: code benchmark evaluation and eval plan (Refs ALB-006)

noahgift
and
claude
committed
fix: Use quantize_q4_k_matrix for row-aligned Q4K blocks (Refs #370)

noahgift
and
claude
committed
Add --plan flag to merge, export, and publish commands

noahgift
and
claude
committed
feat: add apr pipeline plan/apply/status/validate wrapping forjar (Refs ALB-028)

noahgift
and
claude
committed
feat: Wire --task instruct into apr finetune CLI (Refs GH-371)

noahgift
and
claude
committed
feat: add causal LM pre-training to apr train plan/apply (Refs ALB-009)

noahgift
and
claude
committed
fix: APR GPU path uses fused Q4K kernels for 448x speedup (Refs GH-87)

noahgift
and
claude
committed
feat(distill): add config-driven two-stage distillation workflow (ALB-011)

noahgift
and
claude
committed
feat: enable entrenar parquet feature for Parquet→LMBatch loading (ALB-007)

noahgift
and
claude
committed

Commits on Mar 2, 2026

Commits on Mar 3, 2026

Commits on Mar 4, 2026

Commits on Mar 5, 2026

feat: wire realizar inference in apr compile codegen (Refs #412)

noahgift
and
claude
committed
ci: remove 8 legacy workflow(s) superseded by clean-room gate

Noah Gift
and
claude
committed
ci: fix concurrency group to prevent container collisions

Noah Gift
and
claude
committed
style: cargo fmt (unified CI pipeline prep) (#413)

authored
ci: unified CI pipeline — lint + build + test + pmat gates (#414)

authored
fix: update deny.toml for cargo-deny 0.19 compatibility
noahgift
committed
fix: update bytes 1.11.0 → 1.11.1 (RUSTSEC-2026-0007)
Noah Gift
committed
fix: update clippy lint name for Rust 1.93

Noah Gift
and
claude
committed
feat: Wire `apr data decontaminate` via alimentar quality module (Refs GH-11)

Noah Gift
and
claude
committed

Commits on Mar 6, 2026

Commits on Mar 7, 2026

Commits on Mar 9, 2026

Commits on Mar 10, 2026

Commits on Mar 12, 2026

Commits on Mar 13, 2026

fix: ALB-107 duplicate metadata keys + APR eval format support (Refs #475)

noahgift
and
claude
committed

Commits on Mar 14, 2026

Commits on Mar 15, 2026

Commits on Mar 16, 2026

Commits on Mar 17, 2026

Commits on Mar 18, 2026

Commits on Mar 19, 2026

fix: partial pv lint fixes — training-pipeline kani + tensor-layout if_fails (Refs PMAT-149)

noahgift
and
claude
committed

Commits on Mar 20, 2026

Merge branch 'main' into fix/transpose-performance-388
Aakash-Naidu
authored