mirror of
https://github.com/SillyTavern/SillyTavern.git
synced 2024-12-12 09:26:33 +01:00
59100a0076
I didn't notice that the prompt wording implies impersonation, which is incorrect for common usage. Changed the order of tags used to generate the character's response to the user, rather than the vice versa.
11 lines
317 B
JSON
11 lines
317 B
JSON
{
|
|
"name": "Llama 2",
|
|
"system_prompt": "Write {{char}}'s next reply in this fictional roleplay with {{user}}.\n<</SYS>>\n",
|
|
"system_sequence": "[INST] <<SYS>>\n",
|
|
"stop_sequence": "",
|
|
"input_sequence": "[INST]",
|
|
"output_sequence": "[/INST]",
|
|
"separator_sequence": "\n",
|
|
"wrap": false
|
|
}
|