mannix/
llama3-gradient:70b-instruct-1048k-iq2_s

172 1 year ago

This model extends LLama-3 70B's context length from 8k to over 1m tokens. [I-Quants]

... /
params
577073ffcc6c · 110B
{
"num_keep": 24,
"stop": [
"<|start_header_id|>",
"<|end_header_id|>",
"<|eot_id|>"
]
}