eas/
codellama:34b-16k

260 2 years ago

Codelama with 16k context unlocked in modelfile

2 years ago

be95e5b84e7f · 16GB ·

model
llama
·
33.7B
·
Q3_K_M
params
{ "num_ctx": 16384, "num_gqa": 8, "rope_frequency_base": 1000000, "stop": [
template
[INST] <<SYS>>{{ .System }}<</SYS>> {{ .Prompt }} [/INST]

Readme

13b version is q4_K_M quantization

34b version is q3_K_M quantization