-
-
Notifications
You must be signed in to change notification settings - Fork 477
Open
Description
For the moment I defined a max_tokens equals to 30000 but it could be useful to know the exact max_tokens
const res = await fetch(process.env.OVH_AI_ENDPOINTS_MODEL_URL, {
method: 'POST',
headers: {
'Authorization': `Bearer ${process.env.OVH_AI_ENDPOINTS_ACCESS_TOKEN}`,
'Content-Type': 'application/json'
},
body: JSON.stringify({
model: process.env.OVH_AI_ENDPOINTS_MODEL_NAME,
temperature: 0.0,
top_P: 1.0,
messages: [
{
//System prompt adds more stability to the inferences
role: 'system', content: systemPrompt,
role: 'user', content: userPrompt
}
],
max_tokens: 30000
})
});
Our user prompt and system prompt are not "tiny" so it's necessary to count how many tokens we are sending and then calculate 32000 - this number of tokens.
Use for example: https://www.npmjs.com/package/tiktoken
Metadata
Metadata
Assignees
Labels
No labels