Skip to content

ONNX: Fix FP8 quantization for the second MLP in LayerNormMLP #2577

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

Open
victoroliv2 wants to merge 1 commit into
base: main
Choose a base branch
from
+12 −3
Open

ONNX: Fix FP8 quantization for the second MLP in LayerNormMLP #2577

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
15 changes: 12 additions & 3 deletions transformer_engine/pytorch/module/layernorm_mlp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2243,14 +2243,23 @@ def onnx_forward(

assert not TEDebugState.debug_enabled, "Debug mode is not supported in ONNX export"
assert_warmed_up(self)

# Get quantizers
(
fc1_input_quantizer,
fc1_weight_quantizer,
_,
_,
_,
_,
fc2_input_quantizer,
fc2_weight_quantizer,
output_quantizer,
*_,
fc2_output_quantizer,
_,
_,
_,
) = self._get_quantizers(False, is_grad_enabled)

inp_dtype = inp.dtype

fc1_weight, fc2_weight = self._get_weight_tensors()
Expand Down Expand Up @@ -2324,7 +2333,7 @@ def _clamped_swiglu(x, limit, alpha):

fc2_out = onnx_gemm(fc2_weight, act_out, fc2_bias)

if output_quantizer is not None:
if fc2_output_quantizer is not None:
raise NotImplementedError("ONNX export of quantized output is not supported")

if self.return_layernorm_output:
Expand Down