tt-metal
cpp Easytenstorrent/tt-metal
1,391 stars
397 forks
61 open issues
Active Mar 2026
Beginner-Friendly Issues 61
Issues tagged for new contributors
Fix mish kernel that hit unpack/pack hw configuration LLK_ASSERT-s
#41015 · Mar 30, 2026
`paged_update_cache` kernel doesn't support batch sizes exceeding physical core count
#41010 · Mar 30, 2026
[Models CI] Migrate phase 1 models to new 3-Tier pipelines
#41002 · Mar 30, 2026
Host side Tensor ops should not fail silently
#40993 · Mar 30, 2026
Whisper: decode loop off-by-one in generation_start corrupts KV cache after 2CQ refactor
#40969 · Mar 28, 2026
[area] PI0 LIBERO demo crashes with IndexError when wrist images are missing
#40967 · Mar 28, 2026
bug community
[area] PI0 demo scripts add wrong repo root to sys.path (cwd-dependent import failure)
#40966 · Mar 28, 2026
bug community
Unbounded busy-loop in test_basic_fabric_mux.cpp
#40944 · Mar 27, 2026
Offline accuracy evaluation of VLA models against CPU reference
#40941 · Mar 27, 2026
ttnn.empty_like op misses to copy TensorTopology
#40927 · Mar 27, 2026
bug
concatenate_heads hangs with height-sharded L1 input and DRAM interleaved output
#40925 · Mar 27, 2026
ttnn.group_norm automatic core grid setting
#40916 · Mar 27, 2026
P0 forge
ttnn.group_norm hangs or returns invalid core grids in certain configurations
#40912 · Mar 27, 2026
Tilize test in TTNN for int8 format has strided input on Blackhole.
#40904 · Mar 27, 2026
bug
ttnn.pad segfaults with interleaved TILE_LAYOUT input and sharded output memory_config
#40898 · Mar 27, 2026
op_cat: TM forge data_movement
[ops]: llama_shapes_sharded_writer.cpp missing noc_async_atomic_barrier
#40879 · Mar 27, 2026
bug
bug community
bug
[ops]: rotary_embedding_llama_fused_qk fails under watcher
#40874 · Mar 27, 2026
bug
Pre-sharded weight loading across PCIe endpoints
#40871 · Mar 27, 2026
community
Recommended kernel variants for multi-chip mesh-device graphs
#40870 · Mar 27, 2026
community
[CI Failure] {{ Deep Seek Test }} - {{ Final Op test }}
#40862 · Mar 26, 2026
ci-bug
ttnn.min produces incorrect results for certain tensor shapes and dimensions
#40854 · Mar 26, 2026
P1 op_cat: reduces ign-ops
Layout / Data Type Transformation Disabled in tt-sim
#40850 · Mar 26, 2026
ttnn.linear is slower on L1 than DRAM
#40845 · Mar 26, 2026
[area]: 2 consecutive max_pool2d ops give wrong results
#40833 · Mar 26, 2026
bug
Migration Datapath Performance Evals
#40831 · Mar 26, 2026
[QSR] Fix Qsr tests that have 32bit dest and DstSync::Half
#40827 · Mar 26, 2026
LLK Quasar
Add support of logprobs>1 for other models
#40810 · Mar 26, 2026
bug P2 llms_on_metal
Clean up CI Model weights cache
#40799 · Mar 26, 2026
P2 models-ci Models Pipeline
[QSR] Optimize llk unpack/pack block operations by using ntiles per unpack/pack
#40798 · Mar 26, 2026
LLK Quasar
Auto-approve codeowner-bypass for first-party dependency bump PRs (UMD, TTSIM, SFPI)
#40789 · Mar 26, 2026
infra-ci dragonstrike area: infra
Test SP4 specific demo on SP5
#40785 · Mar 25, 2026
New Feature: Greater test coverage in Fabric Unit tests
#40780 · Mar 25, 2026
Fabric
Integrate Prefill + Decode KV Cache Chunk Metadata to Lookup Table
#40779 · Mar 25, 2026
Add getter to return KV Chunk Metadata for Migration
#40775 · Mar 25, 2026
[qwen3-vl]: Decode Performance Improvements for Qwen3-VL-32B on WH T3K
#40765 · Mar 25, 2026
bug
Default-constructed Tensor objects crash during hashing
#40745 · Mar 25, 2026
ttnn
glean CI maintenance
glean CI maintenance
glean CI maintenance
feature
[mHC] Module Wrapper
#40726 · Mar 25, 2026
Shield mHC
[mHC] Multi-Chip
#40723 · Mar 25, 2026
Shield mHC
[mHC] General Multi-Core Kernel
#40722 · Mar 25, 2026
Shield mHC
[mHC] Batch and Prefill
#40721 · Mar 25, 2026
Shield mHC
[mHC] Sharded Input Support (B=1, S=1)
#40720 · Mar 25, 2026
Shield mHC
[mHC] Multi-Core Unit Kernel (B=1, S=1)
#40719 · Mar 25, 2026
Shield mHC
[mHC] Single user decode kernel
#40717 · Mar 25, 2026
Shield mHC
Add AIME24 eval as a CI test
#40713 · Mar 25, 2026
[mHC] Single-Core Unit Kernel (B=1, S=1)
#40707 · Mar 25, 2026
Shield mHC
[mHC] PyTorch Reference Implementation
#40706 · Mar 25, 2026
Shield mHC
[mHC] Prototype
#40704 · Mar 25, 2026
Shield mHC
[DM] Remove non-posted semaphore NOC functions
#40699 · Mar 25, 2026
data_movement good first issue
Allow disabling trace capture in simple_text_demo.py
#40694 · Mar 25, 2026
models
Support int32 dtype for embedding in TTNN
#40693 · Mar 25, 2026
feature community
Investigate flaky tt-triage tests in merge gate
#40659 · Mar 24, 2026
Performance Degradation of Llama3.1-8B model in CI
#40637 · Mar 24, 2026
bug