1.1 KiB
1.1 KiB
InferenceAzureAiStudioTaskSettings [interface-InferenceAzureAiStudioTaskSettings]
| Name | Type | Description |
|---|---|---|
do_sample |
float | For a completion task, instruct the inference process to perform sampling. It has no effect unless temperature or top_p is specified. |
max_new_tokens |
integer | For a completion task, provide a hint for the maximum number of output tokens to be generated. |
temperature |
float | For a completion task, control the apparent creativity of generated completions with a sampling temperature. It must be a number in the range of 0.0 to 2.0. It should not be used if top_p is specified. |
top_p |
float | For a completion task, make the model consider the results of the tokens with nucleus sampling probability. It is an alternative value to temperature and must be a number in the range of 0.0 to 2.0. It should not be used if temperature is specified. |
user |
string | For a text_embedding task, specify the user issuing the request. This information can be used for abuse detection. |