MCPcopy Create free account
hub / github.com/pytorch/executorch / _conv_transpose1d_handler

Function _conv_transpose1d_handler

backends/mlx/ops.py:2621–2646  ·  view source on GitHub ↗

Handle aten.conv_transpose1d: (input, weight, bias, stride, padding, output_padding, groups, dilation).

(P: MLXProgramBuilder, n: Node)

Source from the content-addressed store, hash-verified

2619
2620@REGISTRY.register(target=[torch.ops.aten.conv_transpose1d.default])
2621def _conv_transpose1d_handler(P: MLXProgramBuilder, n: Node) -> Slot:
2622 """Handle aten.conv_transpose1d: (input, weight, bias, stride, padding, output_padding, groups, dilation)."""
2623 require_args(n.args, 2, 8, "aten.conv_transpose1d")
2624 require_kwargs(P.kwargs(n), set(), "aten.conv_transpose1d")
2625 x_node, w_node = n.args[0:2]
2626 bias_node = n.args[2] if len(n.args) > 2 else None
2627 groups = n.args[6] if len(n.args) > 6 else 1
2628
2629 stride = _normalize_conv_param(n.args[3] if len(n.args) > 3 else 1, 1, 1)
2630 padding = _normalize_conv_param(n.args[4] if len(n.args) > 4 else 0, 1, 0)
2631 output_padding = _normalize_conv_param(n.args[5] if len(n.args) > 5 else 0, 1, 0)
2632 dilation = _normalize_conv_param(n.args[7] if len(n.args) > 7 else 1, 1, 1)
2633
2634 return _emit_conv_transpose(
2635 P,
2636 n,
2637 x_node,
2638 w_node,
2639 bias_node,
2640 stride,
2641 padding,
2642 dilation,
2643 output_padding,
2644 groups,
2645 ndim=1,
2646 )
2647
2648
2649@REGISTRY.register(target=[torch.ops.aten.conv_transpose2d.input])

Callers

nothing calls this directly

Calls 5

require_argsFunction · 0.85
require_kwargsFunction · 0.85
_normalize_conv_paramFunction · 0.85
_emit_conv_transposeFunction · 0.85
kwargsMethod · 0.80

Tested by

no test coverage detected