Issues

• Feature Request: Standalone Training Loop Support for Android on CPU (NPU + GPGPU solutions are discussed too)

  type:feature
  tensorflow/tensorflow#129927 · Filter by author SwuduSusuwu opened on Nov 3, 2026
  12

  

• [Sparse] Support SparseTensorValue in tf.sparse.bincount and validate_sparse_weights

  tensorflow/tensorflow#129921 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  

• [Sparse] Validate input and convert SparseTensorValue in tf.sparse.reduce_sum and tf.sparse.reduce_max

  type:bug
  tensorflow/tensorflow#129916 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  

• tf.sparse.softmax fails on non-SparseTensor inputs and does not convert SparseTensorValue

  type:bug
  tensorflow/tensorflow#129908 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  

• tf.sparse.maximum and tf.sparse.minimum fail on non-SparseTensor inputs and do not convert SparseTensorValue

  type:bug
  tensorflow/tensorflow#129904 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  

• tf.sparse.expand_dims fails on SparseTensorValue and non-SparseTensor inputs

  type:bug
  tensorflow/tensorflow#129901 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  

• tf.sparse.transpose fails on SparseTensorValue and non-SparseTensor inputs

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129897 · Filter by author Hrushikesh-ramilla opened on Nov 3, 2026
  1

  

• tf.fill: documented NotFoundError for non-integer dims is never raised; a float tensor entry is silently truncated

  2.21.0awaiting PR mergecomp:opstype:bugtype:docs-bug
  tensorflow/tensorflow#129880 · Filter by author Redempt1onzzZZ opened on Nov 3, 2026
  1

  

• tf.math.igammac / tf.math.igamma: docs list bfloat16 and half, but there is no kernel for them on CPU or GPU

  2.21.0awaiting PR mergecomp:opstype:docs-bug
  tensorflow/tensorflow#129878 · Filter by author Redempt1onzzZZ opened on Nov 3, 2026
  

• tf.signal rfft/irfft family and fftnd/ifftnd/rfftnd/irfftnd: Args document input, but the parameter is input_tensor

  2.21.0awaiting PR mergecomp:signaltype:docs-bug
  tensorflow/tensorflow#129875 · Filter by author Redempt1onzzZZ opened on Nov 3, 2026
  

• CsvDataset loses rows when restoring an iterator from a checkpoint with header=True

  2.21.0awaiting PR mergecomp:datatype:bug
  tensorflow/tensorflow#129842 · Filter by author bodapatisaikrishna opened on Nov 2, 2026
  

• tf.dtypes.saturate_cast: int64 input clamps to INT64_MIN on x86; uint32/uint64 input raises NotFoundError

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129838 · Filter by author bodapatisaikrishna opened on Nov 2, 2026
  

• tf.math.argmax rejects NumPy integers as axis, while tf.math.argmin accepts them

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129821 · Filter by author Phoebus-Liu opened on Nov 2, 2026
  1

  

• tf.image.draw_bounding_boxes raises TypeError when colors argument is omitted in TF 2

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129799 · Filter by author Hrushikesh-ramilla opened on Nov 1, 2026
  1

  

• [Google Play Services - TFLite] GPU delegate unavailable on Android 17: tflite_gpu_dynamite no longer enabled (worked on Android 16, same Play services build)

  comp:litetype:support
  tensorflow/tensorflow#129748 · Filter by author alifesoftware-work opened on Oct 30, 2026
  1

  

• tf.image.adjust_saturation turns red pixels gray or black on CPU when blue is a few ulps above green

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129605 · Filter by author usmanmateen opened on Oct 27, 2026
  1

  

• [CUDA] illegal memory read on MaxPoolV2 with qint8 NCHW_VECT_C

  2.20.0awaiting PR mergecomp:gpucomp:opstype:bug
  tensorflow/tensorflow#129587 · Filter by author kokol16 opened on Oct 27, 2026
  1

  

• [CUDA] illegal memory read on ResourceScatterAdd with float16

  2.20.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129577 · Filter by author kokol16 opened on Oct 26, 2026
  2

  

• [CUDA] illegal memory read on MaxPoolWithArgmax

  2.20.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129576 · Filter by author kokol16 opened on Oct 26, 2026
  1

  

• tf.gather_nd with int16 indices returns uninitialized memory or aborts when indices hold 32,768 or more index tuples

  2.21.0awaiting PR mergecomp:opstype:bug
  tensorflow/tensorflow#129561 · Filter by author joszamama opened on Oct 26, 2026
  1

  

• TF 2.21.0 source build fails: LLVM checksum mismatch "Checksum was 00b1077e… but wanted 3f986184…"

  2.21.0awaiting PR mergesubtype: ubuntu/linuxtype:build/install
  tensorflow/tensorflow#129543 · Filter by author KEERTHANA-G-R opened on Oct 26, 2026
  1

  

• float32 tf.matmul on rank-3 tensors returns wrong, nondeterministic results on CPU (oneDNN) for batch 2-16 and contraction >= 20000

  2.21.0awaiting PR mergecomp:coretype:bug
  tensorflow/tensorflow#129498 · Filter by author italo-goncalves opened on Oct 23, 2026
  3

  

• overlap_and_add returns non-empty tensor for 0-frame inputs when frame_length > frame_step

  comp:signalstat:awaiting responseTF 2.19type:bug
  tensorflow/tensorflow#129484 · Filter by author Abhirup0 opened on Oct 23, 2026
  2

  

• tf.get_static_value() never folds StridedSlice, so tf.shape(x)[0] has no static value in graph mode

  2.21.0awaiting PR mergecomp:coretype:bug
  tensorflow/tensorflow#129438 · Filter by author Phoebus-Liu opened on Oct 22, 2026
  1

  

• tf.raw_ops.For aborts on GPU for a valid float32 loop while the CPU execution succeeds

  2.21.0comp:gpucomp:opsstat:awaiting tensorflowertype:bug
  tensorflow/tensorflow#129396 · Filter by author lv20031001 opened on Oct 21, 2026
  1