Commit 58bee01b authored by yuguo-Jack's avatar yuguo-Jack
Browse files

readme

parent cc5c9409
...@@ -236,6 +236,8 @@ python3 -u -m paddle.distributed.launch --gpus "0,1,2,3,4,5,6,7" finetune_gener ...@@ -236,6 +236,8 @@ python3 -u -m paddle.distributed.launch --gpus "0,1,2,3,4,5,6,7" finetune_gener
### 训练benchmark ### 训练benchmark
使用的GPGPU:8张DCU-Z100L-32G。
数据集使用[tatsu-lab/alpaca · Datasets at Hugging Face](https://huggingface.co/datasets/tatsu-lab/alpaca),将数据集放置在./examples/benchmark/peft/paddle下: 数据集使用[tatsu-lab/alpaca · Datasets at Hugging Face](https://huggingface.co/datasets/tatsu-lab/alpaca),将数据集放置在./examples/benchmark/peft/paddle下:
``` ```
...@@ -256,6 +258,8 @@ RCCL_NCHANNELS=8 HSA_FORCE_FINE_GRAIN_PCIE=1 python3 -m paddle.distributed.launc ...@@ -256,6 +258,8 @@ RCCL_NCHANNELS=8 HSA_FORCE_FINE_GRAIN_PCIE=1 python3 -m paddle.distributed.launc
### 推理benchmark 1 ### 推理benchmark 1
使用的GPGPU:1张DCU-Z100L-32G。
``` ```
cd ./examples/benchmark/peft/paddle cd ./examples/benchmark/peft/paddle
python3 inference_benchmark.py --model_name_or_path facebook/llama-13b --dtype float16 --do_forward --do_generate python3 inference_benchmark.py --model_name_or_path facebook/llama-13b --dtype float16 --do_forward --do_generate
...@@ -263,6 +267,8 @@ python3 inference_benchmark.py --model_name_or_path facebook/llama-13b --dtyp ...@@ -263,6 +267,8 @@ python3 inference_benchmark.py --model_name_or_path facebook/llama-13b --dtyp
### 推理benchmark 2(换用[PaddleNLP-develop](https://github.com/PaddlePaddle/PaddleNLP/tree/28158b9735837495e6c73f848925e1d72b821863)) ### 推理benchmark 2(换用[PaddleNLP-develop](https://github.com/PaddlePaddle/PaddleNLP/tree/28158b9735837495e6c73f848925e1d72b821863))
使用的GPGPU:1张DCU-Z100L-32G。
``` ```
pip3 uninstall paddlenlp pip3 uninstall paddlenlp
cd ./llm cd ./llm
...@@ -272,6 +278,8 @@ python3 predictor.py --model_name_or_path facebook/llama-13b --dtype float16 --s ...@@ -272,6 +278,8 @@ python3 predictor.py --model_name_or_path facebook/llama-13b --dtype float16 --s
### LAMBADA推理评估 ### LAMBADA推理评估
使用的GPGPU:1张DCU-Z100L-32G。
``` ```
cd ./examples/benchmark/lambada cd ./examples/benchmark/lambada
wget https://paddlenlp.bj.bcebos.com/data/benchmark/lambada_test.jsonl wget https://paddlenlp.bj.bcebos.com/data/benchmark/lambada_test.jsonl
......
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