Skip to content

Migrate Gpt3 to NNX. #2062

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
wants to merge 1 commit into
base: main
Choose a base branch
from
+208 −170
Open

Migrate Gpt3 to NNX. #2062

Show file tree
Hide file tree
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
8 changes: 3 additions & 5 deletions src/MaxText/layers/decoders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -411,7 +411,7 @@ def get_decoder_layers(self):
case DecoderBlockType.GEMMA3:
return [gemma3.Gemma3DecoderLayerToLinen]
case DecoderBlockType.GPT3:
return [gpt3.Gpt3DecoderLayer]
return [gpt3.Gpt3DecoderLayerToLinen]
case DecoderBlockType.GPT_OSS:
return [gpt_oss.GptOssScannableBlockToLinen] if self.config.scan_layers else [gpt_oss.GptOssDecoderLayerToLinen]
case DecoderBlockType.QWEN3:
Expand Down Expand Up @@ -590,7 +590,7 @@ def _apply_embedding(
name="position_embedder",
config=cfg,
mesh=self.mesh,
)(decoder_positions, model_mode=model_mode)
)(decoder_positions.astype("int32"), model_mode=model_mode)
return y

@nn.compact
Expand Down Expand Up @@ -843,9 +843,7 @@ def __call__(
# Iterate over the two layer groups (dense and MoE) and apply layer transformation
for layer, num_layers, layer_prefix in zip(layers, num_layers_list, layer_prefixes):
for index in range(num_layers):
y = layer(
config=cfg, mesh=mesh, name=f"{layer_prefix}_{index}", quant=self.quant, model_mode=self.model_mode
)(
y = layer(config=cfg, mesh=mesh, name=f"{layer_prefix}_{index}", quant=self.quant, model_mode=self.model_mode)(
y,
decoder_segment_ids,
decoder_positions,
Expand Down
Loading
Loading