ttnn.kv_cache.fill_cache_for_user_

ttnn.kv_cache.fill_cache_for_user_(cache: ttnn.Tensor, input_tensor: ttnn.Tensor, batch_index: int) → ttnn.Tensor

Populates the cache tensor in-place with values sourced from input at batch_index.

Parameters:

• cache (ttnn.Tensor) – the cache tensor to be written to.

• input_tensor (ttnn.Tensor) – the input tensor to be written to the cache.

• batch_index (int) – the index into the cache tensor.

Returns:

ttnn.Tensor – the output tensor.