Add some results
This commit is contained in:
29
README.md
29
README.md
@ -120,7 +120,7 @@ LawGPT 系列模型的训练过程分为两个阶段:
|
||||
### 二次训练流程
|
||||
|
||||
1. 参考 `data/example_instruction_train.json` 构造指令微调数据集
|
||||
2. 运行 `src/scripts/train.sh`
|
||||
2. 运行 `src/scripts/train_lora.sh`
|
||||
|
||||
### 指令精调步骤
|
||||
|
||||
@ -135,9 +135,21 @@ LawGPT 系列模型的训练过程分为两个阶段:
|
||||
|
||||
### 输出示例
|
||||
|
||||
<details><summary>问题:欠了信用卡的钱还不上要坐牢吗?</summary>
|
||||
<details><summary>问题:请给出判决意见。</summary>
|
||||
|
||||

|
||||

|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>问题:请介绍赌博罪的定义。</summary>
|
||||
|
||||

|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>问题:请问加班工资怎么算?</summary>
|
||||
|
||||

|
||||
|
||||
</details>
|
||||
|
||||
@ -147,12 +159,19 @@ LawGPT 系列模型的训练过程分为两个阶段:
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>问题:你能否写一段抢劫罪罪名的案情描述?</summary>
|
||||
<details><summary>问题:欠了信用卡的钱还不上要坐牢吗?</summary>
|
||||
|
||||

|
||||

|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>问题:你能否写一段抢劫罪罪名的案情描述?</summary>
|
||||
|
||||

|
||||
|
||||
</details>
|
||||
|
||||
|
||||
### 局限性
|
||||
|
||||
由于计算资源、数据规模等因素限制,当前阶段 LawGPT 存在诸多局限性:
|
||||
|
||||
|
Before Width: | Height: | Size: 605 KiB After Width: | Height: | Size: 605 KiB |
BIN
assets/demo/example-04.jpeg
Normal file
BIN
assets/demo/example-04.jpeg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 592 KiB |
BIN
assets/demo/example-05.jpeg
Normal file
BIN
assets/demo/example-05.jpeg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 584 KiB |
BIN
assets/demo/example-06.jpeg
Normal file
BIN
assets/demo/example-06.jpeg
Normal file
Binary file not shown.
|
After Width: | Height: | Size: 332 KiB |
@ -1,6 +1,6 @@
|
||||
#!/bin/bash
|
||||
|
||||
WORLD_SIZE=8 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 torchrun --nproc_per_node=8 --master_port=1235 train.py \
|
||||
WORLD_SIZE=8 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 torchrun --nproc_per_node=8 --master_port=1235 train_lora.py \
|
||||
--base_model '../models/base_models/chinese_llama_7b' \
|
||||
--data_path '' \
|
||||
--output_dir '../models/lora_weights' \
|
||||
|
||||
Reference in New Issue
Block a user