Skip to content

[JAX] Estimate post-RHT amax using regular amax #2479

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
jberchtold-nvidia wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

[JAX] Estimate post-RHT amax using regular amax #2479

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 22 additions & 11 deletions transformer_engine/jax/cpp_extensions/quantization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -807,17 +807,28 @@ def _quantize_dbias_impl(
use_rht = True
rht_matrix = get_rht_matrix()

new_amax, post_rht_amax = calculate_post_rht_amax(
x.data,
amax_scope=amax_scope,
transpose_batch_sequence=transpose_batch_sequence,
produce_regular_amax=amax is None,
flatten_axis=flatten_axis,
)
if amax is None:
# If amax is already calculated in a previous layer, we skip calculating it in the TE kernel
# So here we only calculate and update amax when it is not provided from a previous layer (amax is None)
amax = new_amax
use_approx_amax = True
if use_approx_amax:
if amax is None:
amax = calculate_amax(
x.data,
amax_scope=amax_scope,
transpose_batch_sequence=transpose_batch_sequence,
)
amax_scale = 2.0
post_rht_amax = amax * amax_scale
else:
new_amax, post_rht_amax = calculate_post_rht_amax(
x.data,
amax_scope=amax_scope,
transpose_batch_sequence=transpose_batch_sequence,
produce_regular_amax=amax is None,
flatten_axis=flatten_axis,
)
if amax is None:
# If amax is already calculated in a previous layer, we skip calculating it in the TE kernel
# So here we only calculate and update amax when it is not provided from a previous layer (amax is None)
amax = new_amax

if quantizer.scaling_mode == ScalingMode.CURRENT_TENSOR_SCALING:
if amax is None:
Expand Down
Loading