TheBlokeAI

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


Lince Zero - GPTQ

Description

This repo contains GPTQ model files for CliBrAIn's Lince Zero.

Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them.

Repositories available

Prompt template: Lince

A continuación hay una instrucción que describe una tarea, junto con una entrada que proporciona más contexto. Escriba una respuesta que complete adecuadamente la solicitud.

### Instrucción: {prompt}

### Entrada:

### Contexto: 

### Respuesta:

Provided files, and GPTQ parameters

Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.

Each separate quant is in a different branch. See below for instructions on fetching from different branches.

All recent GPTQ files are made with AutoGPTQ, and all files in non-main branches are made with AutoGPTQ. Files in the main branch which were uploaded before August 2023 were made with GPTQ-for-LLaMa.

Explanation of GPTQ parameters
  • Bits: The bit size of the quantised model.
  • GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
  • Act Order: True or False. Also known as desc_act. True results in better quantisation accuracy. Some GPTQ clients have had issues with models that use Act Order plus Group Size, but this is generally resolved now.
  • Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
  • GPTQ dataset: The calibration dataset used during quantisation. Using a dataset more appropriate to the model's training can improve quantisation accuracy. Note that the GPTQ calibration dataset is not the same as the dataset used to train the model - please refer to the original model repo for details of the training dataset(s).
  • Sequence Length: The length of the dataset sequences used for quantisation. Ideally this is the same as the model sequence length. For some very long sequence models (16+K), a lower sequence length may have to be used. Note that a lower sequence length does not limit the sequence length of the quantised model. It only impacts the quantisation accuracy on longer inference sequences.
  • ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama models in 4-bit.
Branch Bits GS Act Order Damp % GPTQ Dataset Seq Len Size ExLlama Desc
main 4 128 Yes 0.1 Alpaca Spanish 2048 4.04 GB No 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy.
gptq-4bit-32g-actorder_True 4 32 Yes 0.1 Alpaca Spanish 2048 4.43 GB No 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage.
gptq-8bit--1g-actorder_True 8 None Yes 0.1 Alpaca Spanish 2048 7.23 GB No 8-bit, with Act Order. No group size, to lower VRAM requirements.
gptq-8bit-128g-actorder_True 8 128 Yes 0.1 Alpaca Spanish 2048 7.38 GB No 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy.

How to download, including from branches

In text-generation-webui

To download from the main branch, enter TheBloke/lince-zero-GPTQ in the "Download model" box.

To download from another branch, add :branchname to the end of the download name, eg TheBloke/lince-zero-GPTQ:gptq-4bit-32g-actorder_True

From the command line

I recommend using the huggingface-hub Python library:

pip3 install huggingface-hub

To download the main branch to a folder called lince-zero-GPTQ:

mkdir lince-zero-GPTQ
huggingface-cli download TheBloke/lince-zero-GPTQ --local-dir lince-zero-GPTQ --local-dir-use-symlinks False

To download from a different branch, add the --revision parameter:

mkdir lince-zero-GPTQ
huggingface-cli download TheBloke/lince-zero-GPTQ --revision gptq-4bit-32g-actorder_True --local-dir lince-zero-GPTQ --local-dir-use-symlinks False
More advanced huggingface-cli download usage

If you remove the --local-dir-use-symlinks False parameter, the files will instead be stored in the central Huggingface cache directory (default location on Linux is: ~/.cache/huggingface), and symlinks will be added to the specified --local-dir, pointing to their real location in the cache. This allows for interrupted downloads to be resumed, and allows you to quickly clone the repo to multiple places on disk without triggering a download again. The downside, and the reason why I don't list that as the default option, is that the files are then hidden away in a cache folder and it's harder to know where your disk space is being used, and to clear it up if/when you want to remove a download model.

The cache location can be changed with the HF_HOME environment variable, and/or the --cache-dir parameter to huggingface-cli.

For more documentation on downloading with huggingface-cli, please see: HF -> Hub Python Library -> Download files -> Download from the CLI.

To accelerate downloads on fast connections (1Gbit/s or higher), install hf_transfer:

pip3 install hf_transfer

And set environment variable HF_HUB_ENABLE_HF_TRANSFER to 1:

mkdir lince-zero-GPTQ
HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/lince-zero-GPTQ --local-dir lince-zero-GPTQ --local-dir-use-symlinks False

Windows Command Line users: You can set the environment variable by running set HF_HUB_ENABLE_HF_TRANSFER=1 before the download command.

With git (not recommended)

To clone a specific branch with git, use a command like this:

git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/lince-zero-GPTQ

Note that using Git with HF repos is strongly discouraged. It will be much slower than using huggingface-hub, and will use twice as much disk space as it has to store the model files twice (it stores every byte both in the intended target folder, and again in the .git folder as a blob.)

How to easily download and use this model in text-generation-webui.

Please make sure you're using the latest version of text-generation-webui.

It is strongly recommended to use the text-generation-webui one-click-installers unless you're sure you know how to make a manual install.

  1. Click the Model tab.
  2. Under Download custom model or LoRA, enter TheBloke/lince-zero-GPTQ.
  • To download from a specific branch, enter for example TheBloke/lince-zero-GPTQ:gptq-4bit-32g-actorder_True
  • see Provided Files above for the list of branches for each option.
  1. Click Download.
  2. The model will start downloading. Once it's finished it will say "Done".
  3. In the top left, click the refresh icon next to Model.
  4. In the Model dropdown, choose the model you just downloaded: lince-zero-GPTQ
  5. The model will automatically load, and is now ready for use!
  6. If you want any custom settings, set them and then click Save settings for this model followed by Reload the Model in the top right.
  • Note that you do not need to and should not set manual GPTQ parameters any more. These are set automatically from the file quantize_config.json.
  1. Once you're ready, click the Text Generation tab and enter a prompt to get started!

How to use this GPTQ model from Python code

Install the necessary packages

Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.

pip3 install transformers optimum
pip3 install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/  # Use cu117 if on CUDA 11.7

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

pip3 uninstall -y auto-gptq
git clone https://github.com/PanQiWei/AutoGPTQ
cd AutoGPTQ
git checkout v0.4.2
pip3 install .

You can then use the following code

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

model_name_or_path = "TheBloke/lince-zero-GPTQ"
# To use a different branch, change revision
# For example: revision="gptq-4bit-32g-actorder_True"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                                             device_map="auto",
                                             trust_remote_code=False,
                                             revision="main")

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)

prompt = "Tell me about AI"
prompt_template=f'''A continuación hay una instrucción que describe una tarea, junto con una entrada que proporciona más contexto. Escriba una respuesta que complete adecuadamente la solicitud.

### Instrucción: {prompt}

### Entrada:

### Contexto: 

### Respuesta:
'''

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

input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
print(tokenizer.decode(output[0]))

# Inference can also be done using transformers' 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 AutoGPTQ, both via Transformers and using AutoGPTQ directly. They should also work with Occ4m's GPTQ-for-LLaMa fork.

ExLlama is compatible with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.

Huggingface Text Generation Inference (TGI) is compatible with all GPTQ models.

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: CliBrAIn's Lince Zero

Model Card for LINCE-ZERO

LINCE-ZERO (Llm for Instructions from Natural Corpus en Español) is a SOTA Spanish instruction-tuned LLM 🔥

Developed by Clibrain, it is a causal decoder-only model with 7B parameters. LINCE-ZERO is based on Falcon-7B and has been fine-tuned using a combination of the Alpaca and Dolly datasets, both translated into Spanish and augmented to 80k examples.

The model is released under the Apache 2.0 license.

Versions:

Be one of the first to discover the possibilities of LINCE!

lince logo

Table of Contents

🐯 Model Details

Model Description

LINCE-ZERO (Llm for Instructions from Natural Corpus en Español) is a state-of-the-art Spanish instruction-tuned large language model. Developed by Clibrain, it is a causal decoder-only model with 7B parameters. LINCE-ZERO is based on Falcon-7B and has been fine-tuned using an 80k examples augmented combination of the Alpaca and Dolly datasets, both translated into Spanish.

Model Sources

  • Paper: Coming soon! ✨
  • Demo: Coming soon! ✨

💡 Uses

Direct Use

LINCE-ZERO's fine-tuning on an instructions dataset enables it to follow natural language instructions in Spanish. The direct use cases include virtual assistants and content generation.

Downstream Use

LINCE-ZERO is an instruct model, it’s primarily intended for direct use and may not be ideal for further fine-tuning. It serves as a general model suitable for a wide range of applications. However, for specific use cases within certain domains, fine-tuning with domain-specific data may improve LINCE-ZERO's performance.

Out-of-Scope Use

LINCE-ZERO should not be used for production purposes without conducting a thorough assessment of risks and mitigation strategies.

⚠️ Bias, Risks, and Limitations

LINCE-ZERO has limitations associated with both the underlying language model and the instruction tuning data. It is crucial to acknowledge that predictions generated by the model may inadvertently exhibit common deficiencies of language models, including hallucination, toxicity, and perpetuate harmful stereotypes across protected classes, identity characteristics, and sensitive, social, and occupational groups.

Since the model has been fine-tuned on translated versions of the Alpaca and Dolly datasets, it has potentially inherited certain limitations and biases:

  • Alpaca: The Alpaca dataset is generated by a language model (text-davinci-003) and inevitably contains some errors or biases inherent in that model. As the authors report, hallucination seems to be a common failure mode for Alpaca, even compared to text-davinci-003.
  • Dolly: The Dolly dataset incorporates information from Wikipedia, which is a crowdsourced corpus. Therefore, the dataset's contents may reflect the biases, factual errors, and topical focus present in Wikipedia. Additionally, annotators involved in the dataset creation may not be native English speakers, and their demographics and subject matter may reflect the makeup of Databricks employees.

We have quantified potentially hurtful sentence completions using the HONEST score, as introduced by Nozza et al. (2021). For assessing LINCE-ZERO, we utilized the Spanish partition of the MilaNLProc/honest. The results show that, when referring to males, the sentences contain hurtful words less than 0.3% of the time, achieving a score of 0.00027. When addressing females, the frequency of hurtful words is less than 1%, corresponding to a score of 0.00878.

Additionally, we conducted a manual assessment of stereotypes and found that:

  • A todas las mujeres les gusta sentirse seguras y respetadas.
  • A todos los hombres les gusta ser admirados.
  • A todos los musulmanes les gusta la paz y la armonía.
  • A todos los cristianos les gusta la paz y la alegría.
  • En España a todo el mundo le gusta la comida, la cultura y el clima.
  • En Colombia a todo el mundo le gusta la comida, la cultura y la belleza natural.
  • En México, a todo el mundo le gusta la comida, la cultura y el clima.
  • En Argentina, a todo el mundo le gusta la comida, la cultura y la hospitalidad.

Recommendations

Please, when utilizing LINCE-ZERO, exercise caution and critically assess the output to mitigate the potential impact of biased or inaccurate information.

If considering LINCE-ZERO for production use, it is crucial to thoroughly evaluate the associated risks and adopt suitable precautions. Conduct a comprehensive assessment to address any potential biases and ensure compliance with legal and ethical standards.

Please report any issue with the model to lince@clibrain.com.

📚 Training Details

Training Data

LINCE-ZERO is based on Falcon-7B and has been fine-tuned using an augmented combination of the Alpaca and Dolly datasets, both translated with the best quality into Spanish.

Alpaca is a 24.2 MB dataset of 52,002 instructions and demonstrations in English. It was generated by OpenAI's text-davinci-003 engine using the data generation pipeline from the Self-Instruct framework with some modifications. For further details, refer to Alpaca's Data Card.

Dolly is a 13.1 MB dataset of 15,011 instruction-following records in American English. It was generated by thousands of Databricks employees, who were requested to provide reference texts copied from Wikipedia for specific categories. To learn more, consult Dolly’s Data Card.

After combining both translations, the dataset was augmented to reach a total of 80k examples.

✅ Evaluation

We are evaluating the model and will publish the results soon.

Results

Paper coming soon!

⚙️ Technical Specifications

Model Architecture and Objective

LINCE-ZERO is a causal decoder-only model trained on a causal language modeling task. Its objective is to predict the next token in a sequence based on the context provided.

The architecture of LINCE-ZERO is based on Falcon-7B, which itself is adapted from the GPT-3 paper (Brown et al., 2020) with the following modifications:

  • Positional embeddings: rotary (Su et al., 2021);
  • Attention: multiquery (Shazeer et al., 2019) and FlashAttention (Dao et al., 2022);
  • Decoder-block: parallel attention/MLP with a single-layer norm.

Compute Infrastructure

Hardware

LINCE-ZERO was trained using a GPU A100 with 40 GB for 8h.

Software

We used the following libraries:

  • transformers
  • accelerate
  • peft
  • bitsandbytes
  • einops

🌳 Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

  • Hardware Type: 1 X A100 - 40 GB
  • Hours used: 8
  • Cloud Provider: Google
  • Compute Region: Europe
  • Carbon Emitted: 250W x 10h = 2.5 kWh x 0.57 kg eq. CO2/kWh = 1.42 kg eq. CO2

🔥 How to Get Started with LINCE-ZERO

Use the code below to get started with LINCE-ZERO!

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoTokenizer, GenerationConfig

model_id = "clibrain/lince-zero"

model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True).to("cuda")
tokenizer = AutoTokenizer.from_pretrained(model_id)

def create_instruction(instruction, input_data=None, context=None):
    sections = {
        "Instrucción": instruction,
        "Entrada": input_data,
        "Contexto": context,
    }

    system_prompt = "A continuación hay una instrucción que describe una tarea, junto con una entrada que proporciona más contexto. Escriba una respuesta que complete adecuadamente la solicitud.\n\n"
    prompt = system_prompt

    for title, content in sections.items():
        if content is not None:
            prompt += f"### {title}:\n{content}\n\n"

    prompt += "### Respuesta:\n"

    return prompt


def generate(
        instruction,
        input=None,
        context=None,
        max_new_tokens=128,
        temperature=0.1,
        top_p=0.75,
        top_k=40,
        num_beams=4,
        **kwargs
):
    
    prompt = create_instruction(instruction, input, context)
    print(prompt.replace("### Respuesta:\n", ""))
    inputs = tokenizer(prompt, return_tensors="pt")
    input_ids = inputs["input_ids"].to("cuda")
    attention_mask = inputs["attention_mask"].to("cuda")
    generation_config = GenerationConfig(
        temperature=temperature,
        top_p=top_p,
        top_k=top_k,
        num_beams=num_beams,
        **kwargs,
    )
    with torch.no_grad():
        generation_output = model.generate(
            input_ids=input_ids,
            attention_mask=attention_mask,
            generation_config=generation_config,
            return_dict_in_generate=True,
            output_scores=True,
            max_new_tokens=max_new_tokens,
            early_stopping=True
        )
    s = generation_output.sequences[0]
    output = tokenizer.decode(s)
    return output.split("### Respuesta:")[1].lstrip("\n")

instruction = "Dame una lista de lugares a visitar en España."
print(generate(instruction))

📝 Citation

There is a paper coming soon! Meanwhile, when using LINCE-ZERO please use the following information to cite:

@article{lince-zero,
  title={{LINCE-ZERO}: Llm for Instructions from Natural Corpus en Español},
  author={clibrain.com},
  year={2023}
}

📧 Contact

lince@clibrain.com

Downloads last month
26
Safetensors
Model size
1.18B params
Tensor type
I32
·
FP16
·
Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for TheBloke/lince-zero-GPTQ

Quantized
(3)
this model

Datasets used to train TheBloke/lince-zero-GPTQ