cas/
teuken-7b-q4km:latest

172 4 months ago

Quantized from Teuken-7B-instruct-research-v0.4, an instruction-tuned 7B parameter multilingual large language model (LLM) pre-trained with 4T tokens within the research project OpenGPT-X.

... /
params
49cb20b7c0b7 · 106B
{
"num_ctx": 4096,
"repeat_penalty": 1.1,
"stop": [
"</s>"
],
"temperature": 0.7,
"top_k": 50,
"top_p": 0.95
}