mannix/
llama3-12b:q4_1

375 1 year ago

Meta-Llama-3-12B-Instruct is a depth upscaling merge of llama3-8b from M. Labonne

1 year ago

dae37e580f01 · 7.3GB ·

model
llama
·
11.5B
·
Q4_1
params
{ "num_keep": 24, "stop": [ "<|start_header_id|>", "<|end_header_id|>",
template
{{ if .System }}<|start_header_id|>system<|end_header_id|> {{ .System }}<|eot_id|>{{ end }}{{ if .Pr

Readme

Meta-Llama-3-12B-Instruct is a depth upscaling merge of llama3-8b from M. Labonne

Quantized with imatrix using groups_merged.txt

https://huggingface.co/mlabonne/Meta-Llama-3-12B-Instruct