Commit 15324083 authored by Jiming Ruan's avatar Jiming Ruan
Browse files

Fine tuned setting for 2 pass pipeline

parent f41a43a1
......@@ -535,11 +535,8 @@ float rmsnorm2d_fwd(rmsnorm2d_fwd_traits t,
h_traits('x', 'y', 'xs', 'ys', 1, 4, 1, 512, 4, True, False, False, 0, 0),
h_traits('x', 'y', 'xs', 'ys', 1, 4, 1,1024, 2, True, False, False, 0, 0),
h_traits('x', 'y', 'xs', 'ys', 1, 8, 1,1024, 1, True, False, False, 0, 0)],
'big' :[ h_traits('x', 'y', 'xs', 'ys', 1, 1, 1, 1024, 8, True, False, True, 0, 0),
# h_traits('x', 'y', 'xs', 'ys', 1, 4, 1, 256, 4, True, False, True, 0, 0),
# h_traits('x', 'y', 'xs', 'ys', 1, 2, 1,1024, 2, True, False, True, 0, 0),
# h_traits('x', 'y', 'xs', 'ys', 1, 4, 1,1024, 1, True, False, True, 0, 0)
]}
'big' :[ h_traits('x', 'y', 'xs', 'ys', 1, 1, 1,1024, 8, True, False, True, 0, 0),
h_traits('x', 'y', 'xs', 'ys', 1, 20, 1, 256, 1, True, False, True, 0, 0)]}
total_blob = list()
for hs_key in h_trait_dict:
hs = h_trait_dict[hs_key]
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment