This is a model that is assumed to perform well, but may require more testing and user feedback. Be aware, only models featured within the GUI of GPT4All, are curated and officially supported by Nomic. Use at your own risk.
About
- Static quants of https://huggingface.co/Replete-AI/Replete-LLM-Qwen2-7b at commit e356943
- Quantized by ThiloteE with llama.cpp commit e09a800
These quants were created with a customized configuration that have been proven to not cause visible end of string (eos) tokens during inference with GPT4All. The config.json, generation_config.json and tokenizer_config.json differ from the original configuration as can be found in the original model's repository at the time of creation of these quants.
Prompt Template (for GPT4All)
Example System Prompt:
<|im_start|>system
Below is an instruction that describes a task. Write a response that appropriately completes the request.<|im_end|>
Chat Template:
<|im_start|>user
%1<|im_end|>
<|im_start|>assistant
%2<|im_end|>
Context Length
32768
Use a lower value during inference, if you do not have enough RAM or VRAM.
Provided Quants
| Link | Type | Size/GB | Notes |
|---|---|---|---|
| GGUF | Q4_0 | 5.44 | fast, recommended |
About GGUF
If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.
Here is a handy graph by ikawrakow comparing some quant types (lower is better):
And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
Thanks
I thank Mradermacher and TheBloke for Inspiration to this model card and their contributions to open source. Also 3Simplex for lots of help along the way. Shoutout to the GPT4All and llama.cpp communities :-)
Original Model card:
license: apache-2.0 base_model: - Qwen/Qwen2-7B datasets: - Replete-AI/Everything_Instruct_8k_context_filtered tags: - unsloth language: - en
Replete-LLM-Qwen2-7b
Thank you to TensorDock for sponsoring Replete-LLM you can check out their website for cloud compute rental below.
Replete-LLM is Replete-AI's flagship model. We take pride in releasing a fully open-source, low parameter, and competitive AI model that not only surpasses its predecessor Qwen2-7B-Instruct in performance, but also competes with (if not surpasses) other flagship models from closed source like gpt-3.5-turbo, but also open source models such as gemma-2-9b-it and Meta-Llama-3.1-8B-Instruct in terms of overall performance across all fields and categories. You can find the dataset that this model was trained on linked bellow:
Try bartowski's quantizations:
Cant run the model locally? Well then use the huggingface space instead:
Some statistics about the data the model was trained on can be found in the image and details bellow, while a more comprehensive look can be found in the model card for the dataset. (linked above):
Replete-LLM-Qwen2-7b is a versatile model fine-tuned to excel on any imaginable task. The following types of generations were included in the fine-tuning process:
- Science: (General, Physical Reasoning)
- Social Media: (Reddit, Twitter)
- General Knowledge: (Character-Codex), (Famous Quotes), (Steam Video Games), (How-To? Explanations)
- Cooking: (Cooking Preferences, Recipes)
- Writing: (Poetry, Essays, General Writing)
- Medicine: (General Medical Data)
- History: (General Historical Data)
- Law: (Legal Q&A)
- Role-Play: (Couple-RP, Roleplay Conversations)
- News: (News Generation)
- Coding: (3 million rows of coding data in over 100 coding languages)
- Math: (Math data from TIGER-Lab/MathInstruct)
- Function Calling: (Function calling data from "glaiveai/glaive-function-calling-v2")
- General Instruction: (All of teknium/OpenHermes-2.5 fully filtered and uncensored)
Prompt Template: ChatML
<|im_start|>system
{}<|im_end|>
<|im_start|>user
{}<|im_end|>
<|im_start|>assistant
{}
End token (eot_token)
<|endoftext|>
Want to know the secret sause of how this model was made? Find the write up bellow
Continuous Fine-tuning Without Loss Using Lora and Mergekit
https://docs.google.com/document/d/1OjbjU5AOz4Ftn9xHQrX3oFQGhQ6RDUuXQipnQ9gn6tU/edit?usp=sharing
The code to finetune this AI model can be found bellow
https://colab.research.google.com/drive/1vIrqH5uYDQwsJ4-OO3DErvuv4pBgVwk4?usp=sharing
Note this model in particular was finetuned using an h100 using Tensordock.com using the Pytorch OS. In order to use Unsloth code with TensorDock you need to run the following code (Bellow) to reinstall drivers on TensorDock before unsloth works. After running the code bellow, your Virtual Machine will reset, and you will have to SSH back into it. And then you can run the normal unsloth code in order.
# Check Current Size
!df -h /dev/shm
# Increase Size Temporarily
!sudo mount -o remount,size=16G /dev/shm
# Increase Size Permanently
!echo "tmpfs /dev/shm tmpfs defaults,size=16G 0 0" | sudo tee -a /etc/fstab
# Remount /dev/shm
!sudo mount -o remount /dev/shm
# Verify the Changes
!df -h /dev/shm
!nvcc --version
!export TORCH_DISTRIBUTED_DEBUG=DETAIL
!export NCCL_DEBUG=INFO
!python -c "import torch; print(torch.version.cuda)"
!export PATH=/usr/local/cuda/bin:$PATH
!export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
!export NCCL_P2P_LEVEL=NVL
!export NCCL_DEBUG=INFO
!export NCCL_DEBUG_SUBSYS=ALL
!export TORCH_DISTRIBUTED_DEBUG=INFO
!export TORCHELASTIC_ERROR_FILE=/PATH/TO/torcherror.log
!sudo apt-get remove --purge -y '^nvidia-.*'
!sudo apt-get remove --purge -y '^cuda-.*'
!sudo apt-get autoremove -y
!sudo apt-get autoclean -y
!sudo apt-get update -y
!sudo apt-get install -y nvidia-driver-535 cuda-12-1
!sudo add-apt-repository ppa:graphics-drivers/ppa -y
!sudo apt-get update -y
!sudo apt-get update -y
!sudo apt-get install -y software-properties-common
!sudo add-apt-repository ppa:graphics-drivers/ppa -y
!sudo apt-get update -y
!latest_driver=$(apt-cache search '^nvidia-driver-[0-9]' | grep -oP 'nvidia-driver-\K[0-9]+' | sort -n | tail -1) && sudo apt-get install -y nvidia-driver-$latest_driver
!sudo reboot
Join the Replete-Ai discord! We are a great and Loving community!
- Downloads last month
- 1,240