# Fast-Inference with Ctranslate2

Speedup inference while reducing memory by 2x-4x using int8 inference in C++ on CPU or GPU.

quantized version of mosaicml/mpt-7b-instruct

pip install hf-hub-ctranslate2>=2.12.0 ctranslate2>=3.16.0
# from transformers import AutoTokenizer
model_name = "michaelfeil/ct2fast-mpt-7b-instruct"


from hf_hub_ctranslate2 import GeneratorCT2fromHfHub
model = GeneratorCT2fromHfHub(
        # load in int8 on CUDA
        model_name_or_path=model_name,
        device="cuda",
        compute_type="int8_float16",
        # tokenizer=AutoTokenizer.from_pretrained("{ORG}/{NAME}")
)
outputs = model.generate(
    text=["def fibonnaci(", "User: How are you doing? Bot:"],
    max_length=64,
    include_prompt_in_result=False
)
print(outputs)

Checkpoint compatible to ctranslate2>=3.16.0 and hf-hub-ctranslate2>=2.12.0

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

Converted on 2023-06-27 using

ct2-transformers-converter --model mosaicml/mpt-7b-instruct --output_dir ~/tmp-ct2fast-mpt-7b-instruct --force --copy_files tokenizer.json README.md tokenizer_config.json generation_config.json special_tokens_map.json requirements.txt .gitattributes --quantization int8_float16 --trust_remote_code

Licence and other remarks:

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

Original description

MPT-7B-Instruct

MPT-7B-Instruct is a model for short-form instruction following. It is built by finetuning MPT-7B on a dataset derived from the Databricks Dolly-15k and the Anthropic Helpful and Harmless (HH-RLHF) datasets.

This model was trained by MosaicML and follows a modified decoder-only transformer architecture.

Model Date

May 5, 2023

Model License

CC-By-SA-3.0

Documentation

Example Question/Instruction

Longboi24:

What is a quoll?

MPT-7B-Instruct:

A Quoll (pronounced “cool”) is one of Australia’s native carnivorous marsupial mammals, which are also known as macropods or wallabies in other parts around Asia and South America

How to Use

Note: This model requires that trust_remote_code=True be passed to the from_pretrained method. This is because we use a custom model architecture that is not yet part of the transformers package.

It includes options for many training efficiency features such as FlashAttention (Dao et al. 2022), ALiBi, QK LayerNorm, and more.

import transformers
model = transformers.AutoModelForCausalLM.from_pretrained(
  'mosaicml/mpt-7b-instruct',
  trust_remote_code=True
)

Note: This model requires that trust_remote_code=True be passed to the from_pretrained method. This is because we use a custom MPT model architecture that is not yet part of the Hugging Face transformers package. MPT includes options for many training efficiency features such as FlashAttention, ALiBi, QK LayerNorm, and more.

To use the optimized triton implementation of FlashAttention, you can load the model on GPU (cuda:0) with attn_impl='triton' and with bfloat16 precision:

import torch
import transformers

name = 'mosaicml/mpt-7b-instruct'

config = transformers.AutoConfig.from_pretrained(name, trust_remote_code=True)
config.attn_config['attn_impl'] = 'triton'
config.init_device = 'cuda:0' # For fast initialization directly on GPU!

model = transformers.AutoModelForCausalLM.from_pretrained(
  name,
  config=config,
  torch_dtype=torch.bfloat16, # Load model weights in bfloat16
  trust_remote_code=True
)

Although the model was trained with a sequence length of 2048, ALiBi enables users to increase the maximum sequence length during finetuning and/or inference. For example:

import transformers

name = 'mosaicml/mpt-7b-instruct'

config = transformers.AutoConfig.from_pretrained(name, trust_remote_code=True)
config.max_seq_len = 4096 # (input + output) tokens can now be up to 4096

model = transformers.AutoModelForCausalLM.from_pretrained(
  name,
  config=config,
  trust_remote_code=True
)

This model was trained with the EleutherAI/gpt-neox-20b tokenizer.

from transformers import AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neox-20b")

The model can then be used, for example, within a text-generation pipeline.
Note: when running Torch modules in lower precision, it is best practice to use the torch.autocast context manager.

from transformers import pipeline

pipe = pipeline('text-generation', model=model, tokenizer=tokenizer, device='cuda:0')

with torch.autocast('cuda', dtype=torch.bfloat16):
    print(
        pipe('Here is a recipe for vegan banana bread:\n',
            max_new_tokens=100,
            do_sample=True,
            use_cache=True))

Formatting

This model was trained on data formatted in the dolly-15k format:

INSTRUCTION_KEY = "### Instruction:"
RESPONSE_KEY = "### Response:"
INTRO_BLURB = "Below is an instruction that describes a task. Write a response that appropriately completes the request."
PROMPT_FOR_GENERATION_FORMAT = """{intro}
{instruction_key}
{instruction}
{response_key}
""".format(
    intro=INTRO_BLURB,
    instruction_key=INSTRUCTION_KEY,
    instruction="{instruction}",
    response_key=RESPONSE_KEY,
)

example = "James decides to run 3 sprints 3 times a week. He runs 60 meters each sprint. How many total meters does he run a week? Explain before answering."
fmt_ex = PROMPT_FOR_GENERATION_FORMAT.format(instruction=example)

In the above example, fmt_ex is ready to be tokenized and sent through the model.

Model Description

The architecture is a modification of a standard decoder-only transformer.

The model has been modified from a standard transformer in the following ways:

Hyperparameter Value
n_parameters 6.7B
n_layers 32
n_heads 32
d_model 4096
vocab size 50432
sequence length 2048

PreTraining Data

For more details on the pretraining process, see MPT-7B.

The data was tokenized using the EleutherAI/gpt-neox-20b tokenizer.

Training Configuration

This model was trained on 8 A100-40GBs for about 2.3 hours using the MosaicML Platform. The model was trained with sharded data parallelism using FSDP and used the AdamW optimizer.

Limitations and Biases

The following language is modified from EleutherAI's GPT-NeoX-20B

MPT-7B-Instruct can produce factually incorrect output, and should not be relied on to produce factually accurate information. MPT-7B-Instruct was trained on various public datasets. While great efforts have been taken to clean the pretraining data, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Acknowledgements

This model was finetuned by Sam Havens and the MosaicML NLP team

MosaicML Platform

If you're interested in training and deploying your own MPT or LLMs on the MosaicML Platform, sign up here.

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please cosult an attorney before using this model for commercial purposes.

Citation

Please cite this model using the following format:

@online{MosaicML2023Introducing,
    author    = {MosaicML NLP Team},
    title     = {Introducing MPT-7B: A New Standard for Open-Source, Commercially Usable LLMs},
    year      = {2023},
    url       = {www.mosaicml.com/blog/mpt-7b},
    note      = {Accessed: 2023-03-28}, % change this date
    urldate   = {2023-03-28} % change this date
}
Downloads last month
12
Inference Examples
Inference API (serverless) has been turned off for this model.

Dataset used to train michaelfeil/ct2fast-mpt-7b-instruct