update

README.md

@@ -85,7 +85,9 @@ pip install xentropy_cuda_lib-0.1_torch2.1_dtk2310-cp38-cp38-linux_x86_64.whl
 
 ## 数据集
 `openassistant-guanaco`
-- https://huggingface.co/datasets/timdettmers/openassistant-guanaco/tree/main
+
+[huggingface](https://huggingface.co/datasets/timdettmers/openassistant-guanaco/tree/main) | [SCNet](http://113.200.138.88:18080/aidatasets/openassistant-guanaco) 高速通道
+
 
 项目中已提供用于finetune的迷你数据集，数据目录结构如下：
 ```
@@ -96,16 +98,18 @@ timdettmers/
 官网提供的从头训练的数据集如下，完整数据集的预处理参照[`PRETRAIN.md`](./PRETRAIN.md)：
 
 `SlimPajama-627B`
-- https://huggingface.co/datasets/cerebras/SlimPajama-627B
+
+[huggingface](https://huggingface.co/datasets/cerebras/SlimPajama-627B) | [SCNet](http://113.200.138.88:18080/aidatasets/cerebras/SlimPajama-627B) 高速通道
 
 `starcoderdata`
-- https://huggingface.co/datasets/bigcode/starcoderdata
+
+[huggingface](https://huggingface.co/datasets/bigcode/starcoderdata) | [SCNet](http://113.200.138.88:18080/aidatasets/bigcode/starcoderdata) 高速通道
 
 `更多资料可参考源项目的README_origin.md`
 ## 训练
 ### 单机多卡（finetune）
 ```
-# finetune所需预训练权重下载地址（权重较大需到hf下载）：https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-240k-503b
+# finetune所需预训练权重参考`预训练权重部分`
 # 本步骤说明的预训练权重采用503b，请下载后放入目录PY007下面：PY007/TinyLlama-1.1B-intermediate-step-240k-503b
 cd TinyLlama
 sh sft/script.sh # 全参数finetune
@@ -139,6 +143,11 @@ Assistant: Well, I really don't want him to be president because of his position
 `对话问答`
 ### 热点应用行业
 `制造,广媒,金融,能源,医疗,家居,教育`
+
+## 预训练权重
+
+[huggingface](https://huggingface.co/TinyLlama/TinyLlama-1.1B-intermediate-step-240k-503b) | [SCNet](http://113.200.138.88:18080/aimodels/tinyllama/TinyLlama-1.1B-intermediate-step-240k-503b) 高速通道
+
 ## 源码仓库及问题反馈
 - https://developer.hpccube.com/codes/modelzoo/tinyllama_pytorch
 ## 参考资料