Skip to content

Update lora layer source #2489

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: develop
Choose a base branch
from
Open

Update lora layer source #2489

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
6 changes: 3 additions & 3 deletions paddleformers/peft/lora/lora_quantization_layers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,7 +90,7 @@ class QuantizationLoRALinear(QuantizationLoRABaseLinear):
"""
Quantization lora Linear layer.
The code implementation refers to paddlenformers.peft.lora.lora_layers.LoRALinear.
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddlenformers/peft/lora/lora_layers.py
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddleformers/peft/lora/lora_layers.py
Compare to LoRALinear, this class keeps weight in INT8/INT4 with quant scale, and supports
weight_only_linear for input tensor and origin weight(LoRA part still uses fp16/bf16).
"""
Expand Down Expand Up @@ -128,7 +128,7 @@ class ColumnParallelQuantizationLoRALinear(QuantizationLoRABaseLinear):
"""
Quantization lora Linear layer with mp parallelized(column).
The code implementation refers to paddlenformers.peft.lora.lora_layers.ColumnParallelLoRALinear.
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddlenformers/peft/lora/lora_layers.py#L203
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddleformers/peft/lora/lora_layers.py#L203
Compare to ColumnParallelLoRALinear, this class keeps weight in INT8/INT4 with quant scale, and supports
weight_only_linear for input tensor and origin weight(LoRA part still uses fp16/bf16).
"""
Expand Down Expand Up @@ -205,7 +205,7 @@ class RowParallelQuantizationLoRALinear(QuantizationLoRABaseLinear):
"""
Quantization lora Linear layer with mp parallelized(row).
The code implementation refers to paddlenformers.peft.lora.lora_layers.RowParallelLoRALinear.
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddlenformers/peft/lora/lora_layers.py#L99
https://github.com/PaddlePaddle/PaddleFormers/blob/develop/paddleformers/peft/lora/lora_layers.py#L99
Compare to RowParallelLoRALinear, this class keeps weight in INT8/INT4 with quant scale, and supports
weight_only_linear for input tensor and origin weight(LoRA part still uses fp16/bf16).
"""
Expand Down