TheBloke's picture
Update for Transformers GPTQ support
c748f3b
|
raw
history blame
12.8 kB
metadata
inference: false
license: other
TheBlokeAI

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


Ausboss' Llama 30B SuperCOT GPTQ

These files are GPTQ 4bit model files for Ausboss' Llama 30B SuperCOT merged with Kaio Ken's SuperHOT 8K.

It is the result of quantising to 4bit using GPTQ-for-LLaMa.

This is an experimental new GPTQ which offers up to 8K context size

The increased context is tested to work with ExLlama, via the latest release of text-generation-webui.

It has also been tested from Python code using AutoGPTQ, and trust_remote_code=True.

Code credits:

  • Original concept and code for increasing context length: kaiokendev
  • Updated Llama modelling code that includes this automatically via trust_remote_code: emozilla.

Please read carefully below to see how to use it.

NOTE: Using the full 8K context on a 30B model will exceed 24GB VRAM.

GGML versions are not yet provided, as there is not yet support for SuperHOT in llama.cpp. This is being investigated and will hopefully come soon.

Repositories available

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

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

  1. Click the Model tab.
  2. Under Download custom model or LoRA, enter TheBloke/llama-30b-supercot-SuperHOT-8K-GPTQ.
  3. Click Download.
  4. The model will start downloading. Once it's finished it will say "Done"
  5. Untick Autoload the model
  6. In the top left, click the refresh icon next to Model.
  7. In the Model dropdown, choose the model you just downloaded: llama-30b-supercot-SuperHOT-8K-GPTQ
  8. To use the increased context, set the Loader to ExLlama, set max_seq_len to 8192 or 4096, and set compress_pos_emb to 4 for 8192 context, or to 2 for 4096 context.
  9. Now click Save Settings followed by Reload
  10. The model will automatically load, and is now ready for use!
  11. 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 with AutoGPTQ

First make sure you have AutoGPTQ and Einops installed:

pip3 install einops auto-gptq

Then run the following code. Note that in order to get this to work, config.json has been hardcoded to a sequence length of 8192.

If you want to try 4096 instead to reduce VRAM usage, please manually edit config.json to set max_position_embeddings to the value you want.

from transformers import AutoTokenizer, pipeline, logging
from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
import argparse

model_name_or_path = "TheBloke/llama-30b-supercot-SuperHOT-8K-GPTQ"
model_basename = "llama-30b-supercot-superhot-8k-GPTQ-4bit--1g.act.order"

use_triton = False

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

model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
        model_basename=model_basename,
        use_safetensors=True,
        trust_remote_code=True,
        device_map='auto',
        use_triton=use_triton,
        quantize_config=None)

model.seqlen = 8192

# Note: check the prompt template is correct for this model.
prompt = "Tell me about AI"
prompt_template=f'''USER: {prompt}
ASSISTANT:'''

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

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

# Inference can also be done using transformers' pipeline

# Prevent printing spurious transformers error when using pipeline with AutoGPTQ
logging.set_verbosity(logging.CRITICAL)

print("*** Pipeline:")
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.95,
    repetition_penalty=1.15
)

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

Using other UIs: monkey patch

Provided in the repo is llama_rope_scaled_monkey_patch.py, written by @kaiokendev.

It can be theoretically be added to any Python UI or custom code to enable the same result as trust_remote_code=True. I have not tested this, and it should be superseded by using trust_remote_code=True, but I include it for completeness and for interest.

Provided files

llama-30b-supercot-superhot-8k-GPTQ-4bit--1g.act.order.safetensors

This will work with AutoGPTQ, ExLlama, and CUDA versions of GPTQ-for-LLaMa. There are reports of issues with Triton mode of recent GPTQ-for-LLaMa. If you have issues, please use AutoGPTQ instead.

It was created without group_size to lower VRAM requirements, and with --act-order (desc_act) to boost inference accuracy as much as possible.

  • llama-30b-supercot-superhot-8k-GPTQ-4bit--1g.act.order.safetensors
    • Works for use with ExLlama with increased context (4096 or 8192)
    • Works with AutoGPTQ in Python code, including with increased context, if trust_remote_code=True is set.
    • Should work with GPTQ-for-LLaMa in CUDA mode, but unknown if increased context works - TBC. May have issues with GPTQ-for-LLaMa Triton mode.
    • Works with text-generation-webui, including one-click-installers.
    • Parameters: Groupsize = -1. Act Order / desc_act = True.

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!

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: Sam, theTransient, Jonathan Leane, Steven Wood, webtim, Johann-Peter Hartmann, Geoffrey Montalvo, Gabriel Tamborski, Willem Michiel, John Villwock, Derek Yates, Mesiah Bishop, Eugene Pentland, Pieter, Chadd, Stephen Murray, Daniel P. Andersen, terasurfer, Brandon Frisco, Thomas Belote, Sid, Nathan LeClaire, Magnesian, Alps Aficionado, Stanislav Ovsiannikov, Alex, Joseph William Delisle, Nikolai Manek, Michael Davis, Junyu Yang, K, J, Spencer Kim, Stefan Sabev, Olusegun Samson, transmissions 11, Michael Levine, Cory Kujawski, Rainer Wilmers, zynix, Kalila, Luke @flexchar, Ajan Kanaga, Mandus, vamX, Ai Maven, Mano Prime, Matthew Berman, subjectnull, Vitor Caleffi, Clay Pascal, biorpg, alfie_i, 阿明, Jeffrey Morgan, ya boyyy, Raymond Fosdick, knownsqashed, Olakabola, Leonard Tan, ReadyPlayerEmma, Enrico Ros, Dave, Talal Aujan, Illia Dulskyi, Sean Connelly, senxiiz, Artur Olbinski, Elle, Raven Klaugh, Fen Risland, Deep Realms, Imad Khwaja, Fred von Graf, Will Dee, usrbinkat, SuperWojo, Alexandros Triantafyllidis, Swaroop Kallakuri, Dan Guido, John Detwiler, Pedro Madruga, Iucharbius, Viktor Bowallius, Asp the Wyvern, Edmond Seymore, Trenton Dambrowitz, Space Cruiser, Spiking Neurons AB, Pyrater, LangChain4j, Tony Hughes, Kacper Wikieł, Rishabh Srivastava, David Ziegler, Luke Pendergrass, Andrey, Gabriel Puliatti, Lone Striker, Sebastain Graf, Pierre Kircher, Randy H, NimbleBox.ai, Vadim, danny, Deo Leter

Thank you to all my generous patrons and donaters!

And thank you again to a16z for their generous grant.

Original model card: Kaio Ken's SuperHOT 8K

SuperHOT Prototype 2 w/ 8K Context

This is a second prototype of SuperHOT, this time 30B with 8K context and no RLHF, using the same technique described in the github blog. Tests have shown that the model does indeed leverage the extended context at 8K.

You will need to use either the monkeypatch or, if you are already using the monkeypatch, change the scaling factor to 0.25 and the maximum sequence length to 8192

Looking for Merged & Quantized Models?

Training Details

I trained the LoRA with the following configuration:

  • 1200 samples (~400 samples over 2048 sequence length)
  • learning rate of 3e-4
  • 3 epochs
  • The exported modules are:
    • q_proj
    • k_proj
    • v_proj
    • o_proj
    • no bias
  • Rank = 4
  • Alpha = 8
  • no dropout
  • weight decay of 0.1
  • AdamW beta1 of 0.9 and beta2 0.99, epsilon of 1e-5
  • Trained on 4-bit base model

Original model card: Ausboss' Llama 30B SuperCOT

Merge of huggyllama/llama-30b + kaiokendev/SuperCOT-LoRA

Supercot was trained to work with langchain prompting.

Load up locally in my custom LLM notebook that uses the Oobabooga modules to load up models: https://github.com/ausboss/Local-LLM-Langchain

Then you can add cells from of these other notebooks for testing: https://github.com/gkamradt/langchain-tutorials

From Koikendev Lora page

Compatibility

This LoRA is compatible with any 7B, 13B or 30B 4-bit quantized LLaMa model, including ggml quantized converted bins

Prompting

You should prompt the LoRA the same way you would prompt Alpaca or Alpacino:

Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
<instruction>

### Input:
<any additional context. Remove this if it's not neccesary>

### Response:
<make sure to leave a single new-line here for optimal results>

Remember that with lower parameter sizes, the structure of the prompt becomes more important. The same prompt worded differently can give wildly different answers. Consider using the following suggestion suffixes to improve output quality:

  • "Think through this step by step"
  • "Let's think about this logically"
  • "Explain your reasoning"
  • "Provide details to support your answer"
  • "Compare and contrast your answer with alternatives"

Coming Soon

  • Tweet fix for 13B and 7B - lower model sizes seem to be extremely sensitive to hashtags at the end of training data responses, especially at longer cutoffs