musram opened a new issue, #16364:
URL: https://github.com/apache/tvm/issues/16364
How to get something similar to output[:,0,:] if output is of type of
NDArray?
I am using for example
int64_t X = logits_on_device->shape[0];
int64_t Y = logits_on_device->shape[1];
int64_t Z = logits_on_device->shape[2];
DLDataType dtype = logits_on_device->dtype;
// Create a new NDArray for the slice with shape [X, Z]
NDArray next_tokens_logits = NDArray::Empty({X, Z}, dtype,
logits_on_device->device);
// Copy data from 'logits_on_device' to 'slice'
for (int64_t x = 0; x < X; ++x) {
for (int64_t z = 0; z < Z; ++z) {
// Calculate the index in the original and new array
int original_index = (x * Y * Z) + z; // Y is the size of the
second dimension
int new_index = x * Z + z;
// Copy the element
static_cast<float*>(next_tokens_logits->data)[new_index] =
static_cast<float*>(logits_on_device->data)[original_index];
}
}
I have multiple slice so don't think its optimal to do this operation. How
can these ops can be optimized?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
