Fast-Inference with Ctranslate2

Speedup inference by 2x-8x using int8 inference in C++

quantized version of declare-lab/flan-alpaca-xl

pip install hf_hub_ctranslate2>=1.0.0 ctranslate2>=3.13.0

Checkpoint compatible to ctranslate2 and hf-hub-ctranslate2

compute_type=int8_float16 for device="cuda"
compute_type=int8 for device="cpu"

from hf_hub_ctranslate2 import TranslatorCT2fromHfHub, GeneratorCT2fromHfHub

model_name = "michaelfeil/ct2fast-flan-alpaca-xl"
model = TranslatorCT2fromHfHub(
        # load in int8 on CUDA
        model_name_or_path=model_name, 
        device="cuda",
        compute_type="int8_float16"
)
outputs = model.generate(
    text=["How do you call a fast Flan-ingo?", "Translate to german: How are you doing?"],
    min_decoding_length=24,
    max_decoding_length=32,
    max_input_length=512,
    beam_size=5
)
print(outputs)

Licence and other remarks:

This is just a quantized version. Licence conditions are intended to be idential to original huggingface repo.