apache / tvm

44 Pull requests merged by 22 people

• [TFLite] added scalar axis value handling in reduce

  #6970 merged Dec 9, 2020

• [AutoSchedule] Compatibility improvement with XGBoost v1.3.0

  #7069 merged Dec 9, 2020

• [Auto Scheduler][Auto TVM] Fix infer tile size for NHWC winograd for CUDA

  #7068 merged Dec 9, 2020

• [µTVM] Allow for platform-specific malloc in runtime

  #6948 merged Dec 9, 2020

• PyTorch frontend: make type inference incremental

  #6900 merged Dec 9, 2020

• [FIX] Improve error messages and docs

  #7064 merged Dec 9, 2020

• [NVCC] Fix nvcc compile option to be compatible with older cuda

  #7065 merged Dec 9, 2020

• [TOPI] GPU scatter 1D via sorting based approach

  #7056 merged Dec 9, 2020

• [BYOC][TRT] Support batch norm for all ranks <=5, and all axes

  #7026 merged Dec 8, 2020

• [quantization] fix missing ffi binding of relay.attrs.DequantizeAttrs

  #7054 merged Dec 8, 2020

• [LLVM] Support atomic for GPU backend (NVPTX, ROCm)

  #7051 merged Dec 8, 2020

• [Relay][Frontend][Onnx] MaxUnpool Operator

  #7036 merged Dec 8, 2020

• [Auto Scheduler] Add target host to measure record

  #7046 merged Dec 8, 2020

• [DOCS] Document cloudpickle dependency in tutorials

  #7049 merged Dec 8, 2020

• [TOPI] GPU scatter_add using atomic

  #7044 merged Dec 7, 2020

• [ROCm][Auto scheduler] Support Auto scheduler and NHWC convolution on ROCm

  #7038 merged Dec 6, 2020

• [AutoScheduler] Add tips on resuming the search from a log file

  #7039 merged Dec 5, 2020

• [FIX] disable cuda test for argwhere

  #7042 merged Dec 5, 2020

• [AutoScheduler] Remove `max_registers_per_block` in HardwareParams

  #7040 merged Dec 5, 2020

• [Relay][Topi] Fix GPU NMS when return_indices is True

  #7005 merged Dec 5, 2020

• [AutoScheduler] Improve CPU matmul tutorial

  #7037 merged Dec 5, 2020

• [TVMSCRIPT] Attach span information to tir nodes in tvmscript

  #6910 merged Dec 5, 2020

• Implement Keras Conv1D

  #7035 merged Dec 4, 2020

• [TOPI][OP] cuda for argwhere

  #6868 merged Dec 4, 2020

• [GraphRuntime] remove print from GetInputIndex

  #7027 merged Dec 4, 2020

• [CI] Update docs style dependency.

  #7034 merged Dec 4, 2020

• Add version 11.1 in finding CUDA libdevice

  #7033 merged Dec 4, 2020

• [Frontend] Prevent tflite frontend from producing int64 shape/parameters

  #7030 merged Dec 4, 2020

• [Relay][Frontend][Onnx] Add support for Size op in Onnx frontend.

  #7031 merged Dec 4, 2020

• Save PyTorch frontend state in object

  #7023 merged Dec 4, 2020

• [AutoScheduler] Refactor task interface for tuning single operators

  #7028 merged Dec 4, 2020

• [Relay][Pass] Clean up DCE tests in preparation for refactoring.

  #7029 merged Dec 4, 2020

• [Topi] Fix GPU Dynamic Topk by Improving Dynamic Strided Slice in Topi

  #7018 merged Dec 4, 2020

• [AutoScheduler] Misc update to hardware parameter and task scheduler

  #7020 merged Dec 3, 2020

• [CI] Hotfix CI (see #7010)

  #7025 merged Dec 3, 2020

• [RPC] Prefer IPv4 between IPv4 and IPv6

  #7013 merged Dec 3, 2020

• [Diagnostics] Add environment variable for controlling top-level printing and fix issue with pretty printing/parsing roundtrip.

  #6874 merged Dec 3, 2020

• [µTVM] Fix paths in the reference VM tutorial and add vbguest recommendation

  #7015 merged Dec 3, 2020

• [auto_scheduler] Part.1 metal default hardware params

  #7022 merged Dec 3, 2020

• [AutoScheduler] Add a tutorial on auto-scheduling a network for x86 CPU

  #7019 merged Dec 3, 2020

• Fix trt Test

  #7016 merged Dec 3, 2020

• [AutoScheduler] Support layout rewrite for whole networks

  #6987 merged Dec 3, 2020

• [µTVM] Modify reference VMs to support new µTVM demo

  #7001 merged Dec 3, 2020

• [Backend][Verilator] regression tests

  #7000 merged Dec 2, 2020

14 Pull requests opened by 14 people

• Add AdaTune features

  #7021 opened Dec 3, 2020

• [Arith] Simplify cast

  #7045 opened Dec 7, 2020

• [Frontend][TFLite] Densify Op added

  #7048 opened Dec 7, 2020

• Sparse Conv2D for CPU (NCHW)

  #7050 opened Dec 7, 2020

• [auto_scheduler] buffer support, correctness check

  #7053 opened Dec 8, 2020

• Fix shape init for tvmc onnx frontend

  #7055 opened Dec 8, 2020

• #7058 [Tutorial] Import errors in deploy_detection.py and deploy_classification.py

  #7059 opened Dec 8, 2020

• [WIP] Add MLIR Relay Dialect, Translate MLIR HLO to MLIR Relay to Relay

  #7060 opened Dec 8, 2020

• [Frontend][MXNet] Support mode=instance,spatial for l2_normalize

  #7062 opened Dec 8, 2020

• [Relay][Strategy] Allow cuda cross compilation without physical device.

  #7063 opened Dec 8, 2020

• Add autoscheduler support to tvmc

  #7070 opened Dec 9, 2020

• [AutoScheduler] Delete deprecated file auto_schedule.py

  #7071 opened Dec 9, 2020

• [FIX] Remove debugging print statement

  #7072 opened Dec 9, 2020

• Rollback changes to SSA begin/end scope for Store in C codegen

  #7073 opened Dec 9, 2020

9 Issues closed by 3 people

• Ansor Stabilization Todo Items Tracking

  #6133 closed Dec 9, 2020

• [RESULT][VOTE] Apache Graduation

  #6350 closed Dec 9, 2020

• [DISCUSS][RFC] Apache TVM Graduation

  #6299 closed Dec 9, 2020

• [VOTE] Apache TVM Graduation

  #6332 closed Dec 9, 2020

• [Tutorial] Import errors in deploy_detection.py and deploy_classification.py

  #7058 closed Dec 8, 2020

• Error running YOLOv3-tiny example

  #7061 closed Dec 8, 2020

• c++ run slower than python?

  #7052 closed Dec 8, 2020

• is-there-a-horizontal-fusion-demo

  #7047 closed Dec 7, 2020

• [TEST][FLAKY] tests/python/topi/python/test_topi_argwhere

  #7041 closed Dec 5, 2020

4 Issues opened by 4 people

• [QNN] [BYOC] MergeComposite pass breaks on QNN graph

  #7067 opened Dec 9, 2020

• tvm c++ compil error

  #7066 opened Dec 9, 2020

• [Bug] RecursionError: maximum recursion depth exceeded while calling a Python object

  #7057 opened Dec 8, 2020

• [TEST] Show stdout/stderr on segfault in pytest

  #7017 opened Dec 2, 2020

21 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [ONNX] NMS in ONNX

  #6839 commented on Dec 8, 2020 • 14 new comments

• [TEST][FLAKY] test_op_grad_level2.py::test_conv2d_grad.py

  #7010 commented on Dec 3, 2020 • 8 new comments

• Add retry to sockets on EINTR error

  #6953 commented on Dec 7, 2020 • 7 new comments

• [Relay][Topi]Add Sort Op to Relay

  #6978 commented on Dec 9, 2020 • 6 new comments

• [TOPI] sparse_dense Op sparse_data input added

  #6889 commented on Dec 9, 2020 • 4 new comments

• [Performance] Performance regression with int64 indices INDEX_DEFAULT_I64=ON (PR #6143)

  #6691 commented on Dec 4, 2020 • 2 new comments

• [RELAY][BUG]type inference is slow

  #7008 commented on Dec 6, 2020 • 2 new comments

• [BYOC] Added "include_non_call_ops" parameter to AnnotateTarget pass

  #6655 commented on Dec 9, 2020 • 2 new comments

• [Relay][VM] Add support for references.

  #6798 commented on Dec 8, 2020 • 2 new comments

• [TFLite] pack operation extedned with const args

  #6984 commented on Dec 9, 2020 • 2 new comments

• [TFLite] Strided slice handling of shrink_axis_mask improved

  #6998 commented on Dec 9, 2020 • 2 new comments

• [frontend][keras] Add support for TimeDistributed

  #7006 commented on Dec 9, 2020 • 2 new comments

• Race Condition in Cuda SSD Multibox

  #6631 commented on Dec 7, 2020 • 1 new comment

• Cuda NMS argsorts with invalid entries

  #6629 commented on Dec 7, 2020 • 1 new comment

• [VTA][OpenCL] add device_annot support in graphpack

  #6125 commented on Dec 9, 2020 • 1 new comment

• [VTA][OpenCL] intelfocl

  #6126 commented on Dec 9, 2020 • 1 new comment

• [TF frontend] add support for StridedSlice to input a single constant

  #6949 commented on Dec 8, 2020 • 1 new comment

• [TFLite] Bugfix - ensure pad calcalution to be in int32

  #7009 commented on Dec 4, 2020 • 1 new comment

• Support Boundary Checking for Loop Dependent Iterators

  #6596 commented on Dec 8, 2020 • 0 new comments

• [ETHOSN] Add support for default Ethos-N78 configuration.

  #6982 commented on Dec 9, 2020 • 0 new comments

• [CUDA] BF16 support

  #7014 commented on Dec 3, 2020 • 0 new comments