v0.72.0-dev20260602

Note

If you are installing from a release, please refer to the README, INSTALLATION instructions, and any other documentation packaged with the release, not on the main branch. There may be differences between the latest main and the previous release.

The changelog will now follow, showing the changes from last release.

This release was generated by the CI workflow https://github.com/tenstorrent/tt-metal/actions/runs/26791933184

📦 Uncategorized

• gpt-oss: row-sharded eval + 16K/65K OOM fixes
  • PR: #45633
• DeepSeek V3 Prefill - Moving single-chip tests to L2-nightly
  • PR: #45304
• ds_prefill(model_configs) - Add reference configs for GLM 5.1, MiniMax M2.7, GPT-OSS 120B, DeepSeek V4 Flash/Pro, Kimi K2.6
  • PR: #45408
• Add naive FSDP support
  • PR: #44019
• Reimplement ASSERT
  • PR: #45627
• fix reduce_block_max_row uninit bit 11 leak
  • PR: #45518
• [WIP][tt-train] SDPA Forward: F9 multi-tile K/V chunking
  • PR: #44198
• Support indexed ND-sharded KV in ring joint SDPA
  • PR: #45315
• fix precission regresion after #44412
  • PR: #45692
• ci: upgrade actions/cache v4 → v5 (Node.js 24 compat)
  • PR: #45669
• [Bug fix] batch_norm: update running stats after normalization (#41127)
  • PR: #45578
• gemma4: vLLM bridge for hybrid kv-cache-groups + kv-share aliasing
  • PR: #44265
• [Skip ci] Add BH Galaxy DeepSeek decoder sweep to demo SP release
  • PR: #45263
• BSPM-on-SRAM: unify SRAM hot-expert allocation + demo wiring
  • PR: #45556
• [skip ci] Remove overly broad ttnn catch-all owner mapping
  • PR: #45717
• Device 2.0 Semaphore: add relay_unicast / relay_multicast
  • PR: #45038
• [LLK Test Infra]Strategy based refactor of stimuli generator
  • PR: #45209
• [fuser] refactor yaml validator
  • PR: #45013
• Remove validate-metalium-deprecation pre-commit hook
  • PR: #45364
• Add support for fp8 -> bfp8 for tilize op
  • PR: #44307
• [Feature] More stats in jit_load_report.py
  • PR: #45058
• [skip ci] fabric ubench: update T3K golden CSV after Z-link routing fix (#45420)
  • PR: #45726
• [API Cleanup] Naming cleanup of Metal 2.0 API
  • PR: #45598
• Feature: DRAM-core DRISC prefetcher
  • PR: #45169