imone commited on
Commit
2264eb9
1 Parent(s): a736e30

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -4
README.md CHANGED
@@ -54,7 +54,7 @@ pipeline_tag: text-generation
54
  </div>
55
 
56
  <div style="display: flex; justify-content: center; align-items: center">
57
- <p>* Llama-3-Instruct often fails to follow the few-shot templates.</p>
58
  </div>
59
 
60
  <div align="center">
@@ -87,13 +87,15 @@ curl http://localhost:18888/v1/chat/completions \
87
 
88
  ### Conversation templates
89
 
90
- 💡 **Default Mode**: Best for coding, chat and general tasks
 
 
91
 
92
  ```
93
- GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi<|end_of_turn|>GPT4 Correct User: How are you today?<|end_of_turn|>GPT4 Correct Assistant:
94
  ```
95
 
96
- ⚠️ **Notice:** Remember to set `<|end_of_turn|>` as end of generation token.
97
 
98
  The default template is also available as the integrated `tokenizer.chat_template`, which can be used instead of manually specifying the template:
99
 
 
54
  </div>
55
 
56
  <div style="display: flex; justify-content: center; align-items: center">
57
+ <p>* Llama-3-Instruct often fails to follow the few-shot templates. See <a href="https://huggingface.co/openchat/openchat-3.6-8b-20240522/discussions/6">example</a>.</p>
58
  </div>
59
 
60
  <div align="center">
 
87
 
88
  ### Conversation templates
89
 
90
+ 💡 **Default Mode**: Best for coding, chat and general tasks.
91
+
92
+ It's a modified version of the Llama 3 Instruct template, the only difference is role names, which are either `GPT4 Correct User` or `GPT4 Correct Assistant`
93
 
94
  ```
95
+ <|start_header_id|>GPT4 Correct User<|end_header_id|>\n\nHello<|eot_id|><|start_header_id|>GPT4 Correct Assistant<|end_header_id|>\n\nHi<|eot_id|><|start_header_id|>GPT4 Correct User<|end_header_id|>\n\nHow are you today?<|eot_id|><|start_header_id|>GPT4 Correct Assistant<|end_header_id|>\n\n
96
  ```
97
 
98
+ ⚠️ **Notice:** Remember to set `<|eot_id|>` as end of generation token.
99
 
100
  The default template is also available as the integrated `tokenizer.chat_template`, which can be used instead of manually specifying the template:
101