Tarklanse commited on
Commit
f8387d1
1 Parent(s): c3aea48

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -3
README.md CHANGED
@@ -17,12 +17,12 @@ This is a Lora trained on llama2 7B Chat, with its dataset consisting of a large
17
 
18
  ## Training Dataset
19
 
20
- 微調用的資料集由少量個人撰寫與以此為基礎生成的大量AI生成對話內容組成,使用alpaca-format,約7千條instruction、共9MB的大小
21
 
22
  The fine-tuning dataset used consists of a small number of personally written conversations and a large amount of AI-generated dialogue content based on these, utilizing the alpaca-format. It comprises approximately 7,000 instructions in total and has a size of 9MB.
23
 
24
  ## Training
25
- 使用text-generation-webui中的Training工具,在google colab上調用V100,以4bit模式讀取Llama2 7B後以預設參數訓練,訓練總計花費約2小時
26
 
27
  Using the Training tool in the text-generation-webui, calling a V100 on Google Colab, and reading Llama2 7B in 4-bit mode, training was performed with default parameters. The total training time took approximately 2 hours
28
 
@@ -51,4 +51,8 @@ Prompt需要撰寫角色人格、與個性;如有必要則需要另寫對話
51
  ## License
52
  這個lora以CC BY-SA 4.0作為分享
53
 
54
- 整體使用請遵照Meta的社群許可,不要將其作於非法用途或生成不適當的內容
 
 
 
 
 
17
 
18
  ## Training Dataset
19
 
20
+ 微調用的資料集由少量個人撰寫與以此為基礎生成的大量AI生成對話內容組成,使用alpaca-format,約9千條instruction、共12.6MB的大小
21
 
22
  The fine-tuning dataset used consists of a small number of personally written conversations and a large amount of AI-generated dialogue content based on these, utilizing the alpaca-format. It comprises approximately 7,000 instructions in total and has a size of 9MB.
23
 
24
  ## Training
25
+ 使用text-generation-webui中的Training工具,在google colab上調用V100,以4bit模式讀取Llama2 7B後以Lora Rank64,Lora Alpha128,Epochs5,其餘用預設參數訓練,訓練總計花費約5小時
26
 
27
  Using the Training tool in the text-generation-webui, calling a V100 on Google Colab, and reading Llama2 7B in 4-bit mode, training was performed with default parameters. The total training time took approximately 2 hours
28
 
 
51
  ## License
52
  這個lora以CC BY-SA 4.0作為分享
53
 
54
+ 整體使用請遵照Meta的社群許可,不要將其作於非法用途或生成不適當的內容
55
+
56
+
57
+ ##Update note
58
+ 2023/09/13以現有的資料集改為用更高的Lora Rank,Lora alpha與epoch再次訓練