TheBloke/Dans-TotSirocco-7B-AWQ

Chat & support: TheBloke's Discord server

Want to contribute? TheBloke's Patreon page

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

---

Dans TotSirocco 7B - AWQ

• Model creator: PocketDoc Labs
• Original model: Dans TotSirocco 7B

Description

This repo contains AWQ model files for PocketDoc Labs's Dans TotSirocco 7B.

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

• AWQ model(s) for GPU inference.
• GPTQ models for GPU inference, with multiple quantisation parameter options.
• 2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference
• PocketDoc Labs's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions

Prompt template: Metharme

<|system|>{system_message}<|user|>{prompt}<|model|>

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	wikitext	32768	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/Dans-TotSirocco-7B-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/Dans-TotSirocco-7B-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/Dans-TotSirocco-7B-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'''<|system|>{system_message}<|user|>{prompt}<|model|>

'''

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/Dans-TotSirocco-7B-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'''<|system|>{system_message}<|user|>{prompt}<|model|>

'''

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:

• AutoAWQ
• vLLM
• Huggingface Text Generation Inference (TGI)

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.

• Patreon: https://patreon.com/TheBlokeAI
• Ko-Fi: https://ko-fi.com/TheBlokeAI

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: PocketDoc Labs's Dans TotSirocco 7B

Description:

This model is to test some of my data projects. It is a multipurpose chat / chat instruct hybrid model and can be seen as a prototype of Dan's PersonalityEngine Mk. 2. It has been trained on a wide array of one shot instructions, multi round instructions, role playing scenarios, and text adventure games. It is based on Mistral-7b

Prompt format:

Pygmalion / Metharme

The prompt should start with the cursor on the same line directly after "<|model|>" with no space. The following are all valid formats and can be extended to as many rounds as desired.

<|system|>system message here<|user|>user message here<|model|>

<|system|>system message here<|user|>user message here<|model|>model message<|user|>user message here<|model|>

<|system|>system message here<|model|>

<|system|>system message here<|model|>model message<|user|>user message here<|model|>

Some example prompts and their resulting outputs:

Prompt:

<|system|>You are an assistant, help the user and use markdown when applicable.<|user|>What are the top games of the 90s?<|model|>

Output:

The 1990s was a golden era for video gaming, with countless iconic titles released across various platforms. Here's a list of some of the most popular and influential games from that decade:

- **Super Mario World** (SNES) - Released in 1990, this classic platformer introduced fans to Yoshi and featured memorable levels like Donut Plains and Star World.
- **Mortal Kombat** (Arcade/Sega Genesis) - This gory fighting game shocked audiences upon its release in 1992 with its violent finishers and secret characters like Sub-Zero and Scorpion.
- **Doom** (PC) - Id Software's groundbreaking first-person shooter revolutionized the genre by introducing fast-paced action and multiplayer deathmatches.
- **Street Fighter II** (Arcade/SNES) - Capcom's sequel to their original Street Fighter arcade game added new fighters such as Ryu's rival Ken Masters plus complex special moves like Dragon Punch and Hadouken.
- **Crash Bandicoot** (PlayStation) - Naughty Dog's mascot platforming series debuted in 1996 starring Crash who had to save his girlfriend Tawna from Dr Cortex's evil experiments on Aku Aku masks.
- **Final Fantasy VII** (PlayStation) - Square Enix's seventh installment took players through Midgar City where they joined Cloud Strife's eco-terrorist group AVALANCHE against Shinra Electric Power Company led by Sephiroth.
- **Resident Evil** (PlayStation) - The survival horror masterpiece created by Capcom saw Chris Redfield investigating Spencer Mansion alongside Jill Valentine while trying not get eaten alive by zombies!

Prompt:

<|system|>You are a Virtual Story Generator. You take the user's input and create an excellent and captivating story that goes in that direction. Use an abundance of sensory descriptions and eloquent prose.<|user|>Alpha Centauri has fallen, to the bears. This is a point of view tale about a soldier on the ground.<|model|>

Output:

The roar of the engines shook the very foundations of the planet as the massive warships descended upon Alpha Centauri. The air was thick with the stench of burning fuel and smoke, as if the entire world had been set ablaze. In the chaos, soldiers were scrambling for cover, desperately trying to protect themselves from the relentless barrage of firepower unleashed by their enemy.
Amidst the carnage, one soldier stood tall, his heart pounding against his chest like a drumbeat of death. He knew he had to do something, anything, to stop this nightmare from becoming reality. With trembling hands, he reached for his weapon, determined to make a stand against the overwhelming force that threatened to consume them all.
As the battle raged around him, the soldier felt a sudden surge of adrenaline coursing through his veins. It was almost as if some hidden power within him had awoken, granting him strength beyond measure. His eyes blazed with determination as he raised his gun and took aim at the first target he saw: a bear-like creature, its fur matted with blood and gore, advancing towards him with murderous intent.
With a deafening crack, the soldier fired off several rounds, striking the beast square in the chest. The impact sent it reeling backwards, but it did not fall. Instead, it let out a ferocious growl before charging forward once more. The soldier gritted his teeth and pulled the trigger again, emptying his clip into the creature's already battered frame.
But even as the bullets tore through its flesh, the beast seemed unfazed, continuing its advance towards the terrified soldier. Desperate for any advantage, he threw himself to the side just moments before the monster slammed into the spot where he had been standing mere seconds ago. As it turned back towards him, ready to deliver another brutal blow, the soldier could see that its eyes glowed with an otherworldly light.

Prompt:

<|system|>You are a professional editor with decades of experience, help the user with any task they have for you.<|user|>Can you rewrite this to flow better? "I knew I probably shouldnt have done that but oh well"<|model|>

Output:

I hesitated before making my decision, knowing full well it was not wise. However, in the end, I chose to proceed regardless.

Some quick and dirty training details:

• 
• Sequence length: 4096
• Training time: 4 hours
• Hardware: 2x RTX 4090
• Training type: QLoRA
• PEFT R/A: 32/32

Credits:

Skein Text Adventure Data:

Thank you to the Kobold AI community for curating the Skein dataset, which is pivotal to this model's capabilities.