andreasmadsen commited on
Commit
19db9db
1 Parent(s): 7dd8724

Clarify that model 'tokens' are not actual tokens (#367)

Browse files

The userMessageToken, assistantMessageToken, messageEndToken, and
parameters.stop settings in `MODELS` do not have to be a token. They can
be any string.

Files changed (1) hide show
  1. README.md +5 -4
README.md CHANGED
@@ -118,9 +118,9 @@ MODELS=`[
118
  "datasetName": "OpenAssistant/oasst1",
119
  "description": "A good alternative to ChatGPT",
120
  "websiteUrl": "https://open-assistant.io",
121
- "userMessageToken": "<|prompter|>",
122
- "assistantMessageToken": "<|assistant|>",
123
- "messageEndToken": "</s>",
124
  "preprompt": "Below are a series of dialogues between various people and an AI assistant. The AI tries to be helpful, polite, honest, sophisticated, emotionally aware, and humble-but-knowledgeable. The assistant is happy to help with almost anything, and will do its best to understand exactly what is needed. It also tries to avoid giving false or misleading information, and it caveats when it isn't entirely sure about the right answer. That said, the assistant is practical and really does its best, and doesn't let caution get too much in the way of being useful.\n-----\n",
125
  "promptExamples": [
126
  {
@@ -140,7 +140,8 @@ MODELS=`[
140
  "repetition_penalty": 1.2,
141
  "top_k": 50,
142
  "truncate": 1000,
143
- "max_new_tokens": 1024
 
144
  }
145
  }
146
  ]`
 
118
  "datasetName": "OpenAssistant/oasst1",
119
  "description": "A good alternative to ChatGPT",
120
  "websiteUrl": "https://open-assistant.io",
121
+ "userMessageToken": "<|prompter|>", # This does not need to be a token, can be any string
122
+ "assistantMessageToken": "<|assistant|>", # This does not need to be a token, can be any string
123
+ "messageEndToken": "<|endoftext|>", # This does not need to be a token, can be any string
124
  "preprompt": "Below are a series of dialogues between various people and an AI assistant. The AI tries to be helpful, polite, honest, sophisticated, emotionally aware, and humble-but-knowledgeable. The assistant is happy to help with almost anything, and will do its best to understand exactly what is needed. It also tries to avoid giving false or misleading information, and it caveats when it isn't entirely sure about the right answer. That said, the assistant is practical and really does its best, and doesn't let caution get too much in the way of being useful.\n-----\n",
125
  "promptExamples": [
126
  {
 
140
  "repetition_penalty": 1.2,
141
  "top_k": 50,
142
  "truncate": 1000,
143
+ "max_new_tokens": 1024,
144
+ "stop": ["<|endoftext|>"] # This does not need to be tokens, can be any list of strings
145
  }
146
  }
147
  ]`