int8 quantization of
evo-1-8k-base
, 1.1 revision
# using T4 GPU on CoLab
! pip install optimum-quanto==0.2.1 transformers accelerate flash_attn evo-model --upgrade
from transformers import AutoConfig, AutoModelForCausalLM
from optimum.quanto import quantize, freeze, qint8
hf_model_name = 'togethercomputer/evo-1-8k-base'
model_config = AutoConfig.from_pretrained(
hf_model_name,
trust_remote_code=True,
revision='1.1_fix',
)
model_config.use_cache = True
model = AutoModelForCausalLM.from_pretrained(
hf_model_name,
config=model_config,
trust_remote_code=True,
device_map='auto',
revision='1.1_fix',
)
quantize(model, weights=qint8)
freeze(model)
model.save_pretrained("./quant-evo", safe_serialization=False)