Edit model card
TheBlokeAI

TheBloke's LLM work is generously supported by a grant from andreessen horowitz (a16z)


EM German Leo Mistral - AWQ

Description

This repo contains AWQ model files for Jan Philipp Harries's EM German Leo Mistral.

About AWQ

AWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference.

It is also now supported by continuous batching server vLLM, allowing use of Llama AWQ models for high-throughput concurrent inference in multi-user server scenarios.

As of September 25th 2023, preliminary Llama-only AWQ support has also been added to Huggingface Text Generation Inference (TGI).

Note that, at the time of writing, overall throughput is still lower than running vLLM or TGI with unquantised models, however using AWQ enables using much smaller GPUs which can lead to easier deployment and overall cost savings. For example, a 70B model can be run on 1 x 48GB GPU instead of 2 x 80GB.

Repositories available

Prompt template: EmGerman

Du bist ein hilfreicher Assistent. USER: {prompt} ASSISTANT:

Provided files, and AWQ parameters

For my first release of AWQ models, I am releasing 128g models only. I will consider adding 32g as well if there is interest, and once I have done perplexity and evaluation comparisons, but at this time 32g models are still not fully tested with AutoAWQ and vLLM.

Models are released as sharded safetensors files.

Branch Bits GS AWQ Dataset Seq Len Size
main 4 128 German Quad 4096 4.15 GB

Serving this model from vLLM

Documentation on installing and using vLLM can be found here.

Note: at the time of writing, vLLM has not yet done a new release with AWQ support.

If you try the vLLM examples below and get an error about quantization being unrecognised, or other AWQ-related issues, please install vLLM from Github source.

  • When using vLLM as a server, pass the --quantization awq parameter, for example:
python3 python -m vllm.entrypoints.api_server --model TheBloke/em_german_leo_mistral-AWQ --quantization awq --dtype half

When using vLLM from Python code, pass the quantization=awq parameter, for example:

from vllm import LLM, SamplingParams

prompts = [
    "Hello, my name is",
    "The president of the United States is",
    "The capital of France is",
    "The future of AI is",
]
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)

llm = LLM(model="TheBloke/em_german_leo_mistral-AWQ", quantization="awq", dtype="half")

outputs = llm.generate(prompts, sampling_params)

# Print the outputs.
for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")

Serving this model from Text Generation Inference (TGI)

Use TGI version 1.1.0 or later. The official Docker container is: ghcr.io/huggingface/text-generation-inference:1.1.0

Example Docker parameters:

--model-id TheBloke/em_german_leo_mistral-AWQ --port 3000 --quantize awq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096

Example Python code for interfacing with TGI (requires huggingface-hub 0.17.0 or later):

pip3 install huggingface-hub
from huggingface_hub import InferenceClient

endpoint_url = "https://your-endpoint-url-here"

prompt = "Tell me about AI"
prompt_template=f'''Du bist ein hilfreicher Assistent. USER: {prompt} ASSISTANT:

'''

client = InferenceClient(endpoint_url)
response = client.text_generation(prompt,
                                  max_new_tokens=128,
                                  do_sample=True,
                                  temperature=0.7,
                                  top_p=0.95,
                                  top_k=40,
                                  repetition_penalty=1.1)

print(f"Model output: {response}")

How to use this AWQ model from Python code

Install the necessary packages

Requires: AutoAWQ 0.1.1 or later

pip3 install autoawq

If you have problems installing AutoAWQ using the pre-built wheels, install it from source instead:

pip3 uninstall -y autoawq
git clone https://github.com/casper-hansen/AutoAWQ
cd AutoAWQ
pip3 install .

You can then try the following example code

from awq import AutoAWQForCausalLM
from transformers import AutoTokenizer

model_name_or_path = "TheBloke/em_german_leo_mistral-AWQ"

# Load model
model = AutoAWQForCausalLM.from_quantized(model_name_or_path, fuse_layers=True,
                                          trust_remote_code=False, safetensors=True)
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=False)

prompt = "Tell me about AI"
prompt_template=f'''Du bist ein hilfreicher Assistent. USER: {prompt} ASSISTANT:

'''

print("\n\n*** Generate:")

tokens = tokenizer(
    prompt_template,
    return_tensors='pt'
).input_ids.cuda()

# Generate output
generation_output = model.generate(
    tokens,
    do_sample=True,
    temperature=0.7,
    top_p=0.95,
    top_k=40,
    max_new_tokens=512
)

print("Output: ", tokenizer.decode(generation_output[0]))

"""
# Inference should be possible with transformers pipeline as well in future
# But currently this is not yet supported by AutoAWQ (correct as of September 25th 2023)
from transformers import pipeline

print("*** Pipeline:")
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=512,
    do_sample=True,
    temperature=0.7,
    top_p=0.95,
    top_k=40,
    repetition_penalty=1.1
)

print(pipe(prompt_template)[0]['generated_text'])
"""

Compatibility

The files provided are tested to work with:

TGI merged AWQ support on September 25th, 2023: TGI PR #1054. Use the :latest Docker container until the next TGI release is made.

Discord

For further support, and discussions on these models and AI in general, join us at:

TheBloke AI's Discord server

Thanks, and how to contribute

Thanks to the chirper.ai team!

Thanks to Clay from gpus.llm-utils.org!

I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.

If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.

Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.

Special thanks to: Aemon Algiz.

Patreon special mentions: Pierre Kircher, Stanislav Ovsiannikov, Michael Levine, Eugene Pentland, Andrey, 준교 김, Randy H, Fred von Graf, Artur Olbinski, Caitlyn Gatomon, terasurfer, Jeff Scroggin, James Bentley, Vadim, Gabriel Puliatti, Harry Royden McLaughlin, Sean Connelly, Dan Guido, Edmond Seymore, Alicia Loh, subjectnull, AzureBlack, Manuel Alberto Morcote, Thomas Belote, Lone Striker, Chris Smitley, Vitor Caleffi, Johann-Peter Hartmann, Clay Pascal, biorpg, Brandon Frisco, sidney chen, transmissions 11, Pedro Madruga, jinyuan sun, Ajan Kanaga, Emad Mostaque, Trenton Dambrowitz, Jonathan Leane, Iucharbius, usrbinkat, vamX, George Stoitzev, Luke Pendergrass, theTransient, Olakabola, Swaroop Kallakuri, Cap'n Zoog, Brandon Phillips, Michael Dempsey, Nikolai Manek, danny, Matthew Berman, Gabriel Tamborski, alfie_i, Raymond Fosdick, Tom X Nguyen, Raven Klaugh, LangChain4j, Magnesian, Illia Dulskyi, David Ziegler, Mano Prime, Luis Javier Navarrete Lozano, Erik Bjäreholt, 阿明, Nathan Dryer, Alex, Rainer Wilmers, zynix, TL, Joseph William Delisle, John Villwock, Nathan LeClaire, Willem Michiel, Joguhyik, GodLy, OG, Alps Aficionado, Jeffrey Morgan, ReadyPlayerEmma, Tiffany J. Kim, Sebastain Graf, Spencer Kim, Michael Davis, webtim, Talal Aujan, knownsqashed, John Detwiler, Imad Khwaja, Deo Leter, Jerry Meng, Elijah Stavena, Rooh Singh, Pieter, SuperWojo, Alexandros Triantafyllidis, Stephen Murray, Ai Maven, ya boyyy, Enrico Ros, Ken Nordquist, Deep Realms, Nicholas, Spiking Neurons AB, Elle, Will Dee, Jack West, RoA, Luke @flexchar, Viktor Bowallius, Derek Yates, Subspace Studios, jjj, Toran Billups, Asp the Wyvern, Fen Risland, Ilya, NimbleBox.ai, Chadd, Nitin Borwankar, Emre, Mandus, Leonard Tan, Kalila, K, Trailburnt, S_X, Cory Kujawski

Thank you to all my generous patrons and donaters!

And thank you again to a16z for their generous grant.

Original model card: Jan Philipp Harries's EM German Leo Mistral

EM Logo

In our opinion, this is the strongest open 7b model for German-language applications.

Many thanks to the LeoLM team for the publication of a base model that has received continued pretraining with German texts, greatly improving generation capabilities.

Please note that the Mistral architecture is very recent and still not supported by all libraries (e.g. AutoGPTQ). In case of any problems, please try a different format/base model.

Table of Contents

  1. Introduction
  2. Links & Demos
  3. Prompt Format
  4. Example Output
  5. Acknowledgements
  6. Contact
  7. Disclaimer

Introduction

EM German is a Llama2/Mistral/LeoLM-based model family, finetuned on a large dataset of various instructions in German language. The models are optimized for German text, providing proficiency in understanding, generating, and interacting with German language content.

We offer versions based on 7b, 13b and 70b Llama-2, Mistral and LeoLM (Llama-2/Mistral with continued pretraining on German texts) models.

Please find all Informations, Example Outputs, the special RAG prompt format, output examples and eval results for the EM German Model family in our Github Repository. (Deutsche Version)

Links & Demos

Model Links

Should you try only one model version, I strongly recommend the LeoLM Mistral model which offers by far the best combination of performance and computing requirements!

Base Model HF GPTQ GGUF AWQ
Llama2 7b Link Link Link Link
Llama2 13b Link Link Link Link
Llama2 70b Link Link Link Link
Mistral 7b Link Link Link Link
LeoLM 7b Link Link Link tbc
LeoLM 13b soon soon Link tbc
LeoLM Mistral 7b Link soon Link tbc

Notes about the different versions:

See also the comparison of example outputs for a comparison of (7b) model capabilities.

If you get unsatisfying results with one or another EM German version, please try a different (and/or larger) model or version for your usecase.

Demos:

You can use some of the models with free google Colab instances (e.g. the 7bn model in 8bit or the 13b model with GPTQ):

Prompt Format

This model follows the Vicuna format without linebreaks (but should work with linebreaks as well). The format is as follows:

Du bist ein hilfreicher Assistent. USER: <instruction> ASSISTANT:

You can swap the standard system prompt for a better suited one (see below for RAG-tasks).

Example Output

Please find a selection of various example outputs for all 7b versions of EM German here.

Prompt:

Erkläre die Bedeutung von drei wesentlichen Elementen der Bilanzierung und des Accountings nach IFRS für Unternehmen, die auf internationaler Ebene tätig sind.

Antwort EM German Leo Mistral 7b:

Die International Financial Reporting Standards (IFRS) sind eine Reihe von internationalen Rechnungslegungsstandards, die von der International Accounting Standards Board (IASB) entwickelt wurden. Sie sollen dazu beitragen, dass Finanzberichte vergleichbar, zuverlässig und relevant sind.

1. Fair Value: Dieses Konzept ist ein zentraler Bestandteil der IFRS. Es besagt, dass Vermögenswerte und Schulden zum fairen Wert bilanziert werden sollten, was dem Preis entspricht, zu dem sie zwischen Marktteilnehmern gehandelt werden könnten. Dadurch wird sichergestellt, dass das Unternehmen eine genaue Darstellung seiner finanziellen Lage liefert.

2. Konsolidierung: Die IFRS verlangen von Unternehmen, ihre Finanzergebnisse auf konsolidierter Basis darzustellen. Das bedeutet, dass alle Tochtergesellschaften und verbundenen Unternehmen in die Bilanz des Mutterunternehmens aufgenommen werden. Dadurch wird ein klareres Bild von der finanziellen Situation des Konzerns geschaffen.

3. Vollständigkeitsprinzip: Die IFRS verlangen von Unternehmen, alle relevanten Informationen offenzulegen. Dies umfasst sowohl positive als auch negative Faktoren, wie zum Beispiel Gewinne und Verluste, Einnahmen und Ausgaben sowie Risiken und Unsicherheiten. Durch dieses Prinzip wird sichergestellt, dass Investoren, Kreditgeber und andere Stakeholder eine umfassende Einschätzung der finanziellen Lage des Unternehmens vornehmen können.

Prompt:

Ein Mensch atmet durchschnittlich 16 Mal pro Minute. Wie oft atmet er an einem Tag?

Antwort EM German Leo Mistral 7b:

 Zuerst berechnen wir, wie oft eine Person in einer Stunde atmet. Wir wissen, dass eine Stunde 60 Minuten hat, also:

Atemzüge pro Stunde = Atemfrequenz * Minuten pro Stunde
= 16 Atemzüge pro Minute * 60 Minuten
= 960 Atemzüge

Um herauszufinden, wie viele Atemzüge eine Person an einem Tag macht, müssen wir diese stündliche Rate mit der Anzahl der Stunden an einem Tag multiplizieren, also 24:

Atemzüge pro Tag = Atemzüge pro Stunde * Stunden pro Tag
= 960 Atemzüge * 24 Stunden
= 23.040 Atemzüge

Also macht ein durchschnittlicher Mensch etwa 23.040 Atemzüge an einem Tag.

(For more examples, please visit our Github Repository.)

Acknowledgements:

Many thanks to winglian/caseus for his great work on Axolotl which I used to train the EM mdoels. I am also grateful to Jon Durbin and his Airoboros models and code from which I borrowed many ideas and code snippets. Additionally many thanks to Björn Plüster and the LeoLM team for the outstanding pretraining work on LeoLM and last but not least many many thanks to TheBloke for the preparation of quantized versions in all formats under the sun. The 70b model was trained with support of the OVH Cloud Startup Program.

Contact

I you are interested in customized LLMs for business applications, please get in contact with me via my website. I am also always happy about suggestions and feedback.

PS: We are also always interested in support for our startup ellamind, which will offer customized models for business applications in the future (we are currently still in stealth mode). If you use our models for business applications and have advanced needs for specialized capabilities, please get in touch.

Disclaimer:

I am not responsible for the actions of third parties who use this model or the outputs of the model. This model should only be used for research purposes. The original base model license applies and is distributed with the model files.

Downloads last month
44
Safetensors
Model size
1.2B params
Tensor type
I32
·
FP16
·
Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for TheBloke/em_german_leo_mistral-AWQ

Quantized
(4)
this model