Sam Armstrong

multi_head_attention_forward torch frontend function

2

A frontend implementation for torch.nn.functional.multi_head_attention_forward (which is the functional called by torch.nn.MultiHeadAttention). This is required for transpiling at least one of the HuggingFace models. This frontend implementation captures all the...

fix ivy.dropout returning native arrays/tensors

1

Fixes problem where ivy.dropout can return things like tf.EagerTensor instead of ivy.Array. Min example: ``` import ivy if __name__ == '__main__': ivy.set_backend("tensorflow") x = ivy.random_normal(shape=(5, 5, 5)) x = ivy.dropout(x,...

fix torch frontend avg_pool2d & tests

1

Add ceil_mode, count_include_pad, divisor_override usage to avg_pool2d torch frontend. Also fixed the tests so they all pass 🙂 Test fix depends on https://github.com/unifyai/ivy/pull/17698 being merged.

torch._VF frontend functions

2

Frontend functions need to be added for the following torch._VF functions: - `torch._VF.rnn_tanh` - `torch._VF.rnn_relu` - `torch._VF.gru` - `torch._VF.rnn_tanh_cell` - `torch._VF.rnn_relu_cell` - `torch._VF.lstm_cell` - `torch._VF.gru_cell`

before merging: - get `test_add` passing for the keras backend - add keras module to backend

torch_frontend.Tensor.is_neg

2

torch.Tensor.is_neg() needs to be implemented in the torch frontend in order to write a frontend for [torch.resolve_neg](https://pytorch.org/docs/stable/generated/torch.resolve_neg.html). cc https://github.com/ivy-llc/ivy/pull/27362

torch.Tensor.is_conj() needs to be implemented in the torch frontend in order to write a frontend for [torch.resolve_conj](https://pytorch.org/docs/stable/generated/torch.resolve_conj.html#torch-resolve-conj). cc https://github.com/ivy-llc/ivy/pull/26935

I guess we should just pin the latest version like this, right @hmahmood24?