# TODO(b/304751256): Adjust the numpy pin to a single version, when ready
numpy ~= 1.23.5 ; python_version <= "3.11"
numpy ~= 1.26.0 ; python_version >= "3.12"
wheel ~= 0.41.2
h5py >= 3.10.0
lit ~= 17.0.2
opt_einsum == 3.3.0
astunparse == 1.6.3
dill == 0.3.7
astor == 0.7.1
typing_extensions == 4.8.0
gast == 0.4.0
termcolor == 2.3.0
wrapt == 1.16.0
tblib == 2.0.0
ml_dtypes >= 0.4.0, < 0.5.0

# Install tensorboard, and keras

is raised.

@compatibility(numpy)
Similar to [`numpy.einsum`](https://docs.scipy.org/doc/numpy/reference/generated/numpy.einsum.html).

Comparison with `numpy.einsum`:

 * This Op only supports unary and binary forms of `numpy.einsum`.
 * This Op does not support implicit form. (i.e. equations without `->`).

namex==0.0.8 \
    --hash=sha256:32a50f6c565c0bb10aa76298c959507abdc0e850efe085dc38f3440fcb3aa90b \
    --hash=sha256:7ddb6c2bb0e753a311b7590f84f6da659dd0c05e65cb89d519d54c0a250c0487
    # via keras-nightly
numpy==1.26.4 ; python_version >= "3.12" \
    --hash=sha256:03a8c78d01d9781b28a6989f6fa1bb2c4f2d51201cf99d3dd875df6fbd96b23b \
    --hash=sha256:08beddf13648eb95f8d867350f6a018a4be2e5ad54c8d8caed89ebca558b2818 \

namex==0.0.8 \
    --hash=sha256:32a50f6c565c0bb10aa76298c959507abdc0e850efe085dc38f3440fcb3aa90b \
    --hash=sha256:7ddb6c2bb0e753a311b7590f84f6da659dd0c05e65cb89d519d54c0a250c0487
    # via keras-nightly
numpy==1.23.5 ; python_version <= "3.11" \
    --hash=sha256:01dd17cbb340bf0fc23981e52e1d18a9d4050792e8fb8363cecbf066a84b827d \
    --hash=sha256:06005a2ef6014e9956c09ba07654f9837d9e26696a0470e42beedadb78c11b07 \

namex==0.0.8 \
    --hash=sha256:32a50f6c565c0bb10aa76298c959507abdc0e850efe085dc38f3440fcb3aa90b \
    --hash=sha256:7ddb6c2bb0e753a311b7590f84f6da659dd0c05e65cb89d519d54c0a250c0487
    # via keras-nightly
numpy==1.23.5 ; python_version <= "3.11" \
    --hash=sha256:01dd17cbb340bf0fc23981e52e1d18a9d4050792e8fb8363cecbf066a84b827d \
    --hash=sha256:06005a2ef6014e9956c09ba07654f9837d9e26696a0470e42beedadb78c11b07 \

namex==0.0.8 \
    --hash=sha256:32a50f6c565c0bb10aa76298c959507abdc0e850efe085dc38f3440fcb3aa90b \
    --hash=sha256:7ddb6c2bb0e753a311b7590f84f6da659dd0c05e65cb89d519d54c0a250c0487
    # via keras-nightly
numpy==1.23.5 ; python_version <= "3.11" \
    --hash=sha256:01dd17cbb340bf0fc23981e52e1d18a9d4050792e8fb8363cecbf066a84b827d \
    --hash=sha256:06005a2ef6014e9956c09ba07654f9837d9e26696a0470e42beedadb78c11b07 \

    [`tf.experimental.numpy`](https://www.tensorflow.org/api_docs/python/tf/experimental/numpy),
    which is a NumPy-compatible API for writing TF programs. This module
    provides class `ndarray`, which mimics the `ndarray` class in NumPy, and
    wraps an immutable `tf.Tensor` under the hood. A subset of NumPy functions
    (e.g. `numpy.add`) are provided. Their inter-operation with TF facilities is

        return rewriter.notifyMatchFailure(op, "requires ranked output shape");
      }
      auto rank_diff = ranked_output_type.getRank() - input_type.getRank();
      // The tf.BroadcastTo op performs "right-aligned" numpy-style
      // broadcasting.
      broadcast_dimensions = llvm::to_vector<4>(
          llvm::seq<int64_t>(rank_diff, ranked_output_type.getRank()));
    }
    rewriter.replaceOpWithNewOp<DynamicBroadcastInDimOp>(

* ✨ Support Python internal description on Pydantic model's docstring. PR [#3032](https://github.com/tiangolo/fastapi/pull/3032) by [@Kludex](https://github.com/Kludex).
* ✨ Update `ORJSONResponse` to support non `str` keys and serializing Numpy arrays. PR [#3892](https://github.com/tiangolo/fastapi/pull/3892) by [@baby5](https://github.com/baby5).

### Fixes

	p := s.Prog(op)
	p.To.Type = obj.TYPE_BRANCH
	s.Branches = append(s.Branches, Branch{P: p, B: target})
	return p
}

// DebugFriendlySetPosFrom adjusts Pos.IsStmt subject to heuristics
// that reduce "jumpy" line number churn when debugging.
// Spill/fill/copy instructions from the register allocator,
// phi functions, and instructions with a no-pos position
// are examples of instructions that can cause churn.