tenstorrent/tt-metal
:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.
C++Apache-2.0
Issues
- 4
Define offline MeshGraph descriptor
#14012 opened by aliuTT - 1
[Bug Report] Remove restriction that UnpackToDestFormat must be specified for NUM_CIRCULAR_BUFFERS=32 CBs
#14066 opened by kmabeeTT - 0
Handle Inf, Nan at Kernel level for blocked ops
#14077 opened by VirdhatchaniKN - 0
Add N-dimensional slice
#14100 opened by sjameelTT - 0
[Bug Report] ttnn.clip interface does not follow Pytorch
#14099 opened by kevinwuTT - 1
[Bug Report] Inconsistent Values When Converting Torch Tensor to TTNN Tensor with bfloat8_b Type
#14090 opened by ngohoang34 - 1
Fix single-card nightly and perf pipelines breaking because `huggingface-hub` got rid of `cached_download`
#14076 opened by tt-rkim - 0
Need separate directory for CCL unit tests
#14095 opened by Aswinmcw - 0
ttnn.angle low PCC
#14088 opened by amalbasaTT - 0
ttnn.polar low PCC on GS
#14092 opened by amalbasaTT - 0
[Feature Request] Add unit tests BFP8 for ttnn operations
#14087 opened by mrshaw01 - 0
ttnn.tan low PCC
#14086 opened by dveizovicTT - 1
Bug in reading padded block shard tensors to host.
#14081 opened by sankarmanoj-tt - 0
[Bug Report] Build error with glibc>=2.36
#14083 opened by jeongu-moreh - 0
Need support for Flatten layer
#14082 opened by keerthana-r-mcw - 0
Do preprocessing for Conv2d Weights on the device.
#14080 opened by sankarmanoj-tt - 1
LLKs needed for binary ops
#14073 opened by eyonland - 1
[Bug Report] CPU and NPU produce different BFP8_B values from the same BF16 tensor.
#14032 opened by namhyeong-kim - 0
Reshape Op: use `TT_FATAL` in place of `TT_ASSERTS` + test padded shape instead of unpadded
#14013 opened by mywoodstock - 0
Temporarily take tt-metal-ci-vm-112 out of CI Pool
#14067 opened by tt-asaigal - 0
EMBEDDING: misc/test_rotary_embedding.py::test_rotary_embedding_decode mismatching
#14063 opened by ntarafdar - 0
EMBEDDING: misc/test_rotary_embedding.py::test_rotary_embedding_decode_fp32 mismatching
#14064 opened by ntarafdar - 0
EMBEDDING: misc/test_rotary_embedding_llama.py requires eth connected devices
#14062 opened by ntarafdar - 0
Tilize/Untilize: misc/test_untilize_test.py
#14061 opened by ntarafdar - 0
- 0
[Feature Request] Linear backward operation
#14058 opened by rfurko-tt - 0
Move tt_metal back to C++17
#14055 opened by blozano-tt - 4
[Bug Report] Moreh_adamw program is not properly cached
#14010 opened by rfurko-tt - 0
Remove C++20 dependencies from tt_metal
#14052 opened by patrickroberts - 0
- 0
Remove global BUFFER_MAP
#14038 opened by tt-aho - 0
Runtime Error: ttnn conv2d fails in swin model
#14034 opened by Sudharsan-V - 2
Runtime Error: Convolution output shape mismatch issue if we pass input on device
#13995 opened by punithsekar - 0
- 0
Concat tilize defaults to single-core
#14026 opened by sjameelTT - 0
ttnn::tilize doesn't support block and width sharding
#13979 opened by pavlejosipovic - 0
- 1
Fix Package and Release workflow using system default `CMAKE_INSTALL_PREFIX` after recent CMake refactoring by using `build_metal.sh` underneath in `setup.py`
#14022 opened by tt-rkim - 0
softshrink_bw missing golden function
#14020 opened by VirdhatchaniKN - 0
running asan on main fails
#14015 opened by pgkeller - 1
Add write flush for BH to matmul mcast kernels
#13981 opened by bbradelTT - 0
Enable Non-Causal Flash Decode
#14008 opened by caixunshiren - 0
Document Creation - Inventory of Training Material
#14005 opened by bbeggsTT - 0
- 0
CMake cleanup
#14001 opened by afuller-TT - 0
- 0
- 0
release 1.0: make CommandQueue type opaque
#13994 opened by yan-zaretskiy - 0
Yolov11 - Bring up
#13992 opened by keerthana-r-mcw - 0
T3K Nightly Tests Regression and Inconsistent Test List
#13985 opened by SeanNijjar