mrq/bitsandbytes-rocm
1
0
Fork 1
Code Issues 1 Pull Requests Packages Projects Releases Wiki Activity

Added pre and post device call to transform.

Browse Source
This commit is contained in:
Tim Dettmers 2022-08-04 07:28:12 -07:00
parent 320eacb4c2
commit 6101a8fb9f
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
bitsandbytes/functional.py
View File

@ -1214,6 +1214,7 @@ def transform(A, to_order, from_order='row', out=None, transpose=False, state=No
ptrA = get_ptr(A) ptrA = get_ptr(A)
ptrOut = get_ptr(out) ptrOut = get_ptr(out)
is_on_gpu([A, out]) is_on_gpu([A, out])
prev_device = pre_call(A.device)
if to_order == 'col32': if to_order == 'col32':
if transpose: if transpose:
lib.ctransform_row2col32T(get_ptr(A), get_ptr(out), dim1, dim2) lib.ctransform_row2col32T(get_ptr(A), get_ptr(out), dim1, dim2)
@ -1236,8 +1237,7 @@ def transform(A, to_order, from_order='row', out=None, transpose=False, state=No
lib.ctransform_ampere2row(get_ptr(A), get_ptr(out), dim1, dim2) lib.ctransform_ampere2row(get_ptr(A), get_ptr(out), dim1, dim2)
else: else:
raise NotImplementedError(f'Transform function not implemented: From {from_order} to {to_order}') raise NotImplementedError(f'Transform function not implemented: From {from_order} to {to_order}')
post_call(prev_device)
return out, new_state return out, new_state