exllamav2
ChatGPT-AutoExpert
exllamav2 | ChatGPT-AutoExpert | |
---|---|---|
17 | 31 | |
3,010 | 6,460 | |
- | - | |
9.8 | 8.8 | |
3 days ago | 4 months ago | |
Python | JavaScript | |
MIT License | GNU General Public License v3.0 or later |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
exllamav2
- Running Llama3 Locally
-
Mixture-of-Depths: Dynamically allocating compute in transformers
There are already some implementations out there which attempt to accomplish this!
Here's an example: https://github.com/silphendio/sliced_llama
A gist pertaining to said example: https://gist.github.com/silphendio/535cd9c1821aa1290aa10d587...
Here's a discussion about integrating this capability with ExLlama: https://github.com/turboderp/exllamav2/pull/275
And same as above but for llama.cpp: https://github.com/ggerganov/llama.cpp/issues/4718#issuecomm...
-
What do you use to run your models?
Sorry, I'm somewhat familiar with this term (I've seen it as a model loader in Oobabooga), but still not following the correlation here. Are you saying I should instead be using this project in lieu of llama.cpp? Or are you saying that there is, perhaps, an exllamav2 "extension" or similar within llama.cpp that I can use?
-
I just started having problems with the colab again. I get errors and it just stops. Help?
EDIT: I reported the bug to the exllamav2 Github. It's actually already fixed, just not on any current built release.
-
Yi-34B-200K works on a single 3090 with 47K context/4bpw
install exllamav2 from git with pip install git+https://github.com/turboderp/exllamav2.git. Make sure you have flash attention 2 as well.
-
Tested: ExllamaV2's max context on 24gb with 70B low-bpw & speculative sampling performance
Recent releases for exllamav2 brings working fp8 cache support, which I've been very excited to test. This feature doubles the maximum context length you can run with your model, without any visible downsides.
-
Show HN: Phind Model beats GPT-4 at coding, with GPT-3.5 speed and 16k context
Without batching, I was actually thinking that's kind of modest.
ExllamaV2 will get 48 tokens/s on a 4090, which is much slower/cheaper than an H100:
https://github.com/turboderp/exllamav2#performance
I didn't test codellama, but the 3090 TI figures are in the ballpark of my generation speed on a 3090.
-
Guide for Llama2 70b model merging and exllama2 quantization
First, you need the convert.py script from turboderp's Exllama2 repo. You can read all about the convert.py arguments here.
-
LLM Falcon 180B Needs 720GB RAM to Run
> brute aggressive quantization
Cutting edge quantization like ExLlama's EX2 is far from brute force: https://github.com/turboderp/exllamav2#exl2-quantization
> The format allows for mixing quantization levels within a model to achieve any average bitrate between 2 and 8 bits per weight. Moreover, it's possible to apply multiple quantization levels to each linear layer, producing something akin to sparse quantization wherein more important weights (columns) are quantized with more bits. The same remapping trick that lets ExLlama work efficiently with act-order models allows this mixing of formats to happen with little to no impact on performance. Parameter selection is done automatically by quantizing each matrix multiple times, measuring the quantization error (with respect to the chosen calibration data) for each of a number of possible settings, per layer. Finally, a combination is chosen that minimizes the maximum quantization error over the entire model while meeting a target average bitrate.
Llama.cpp is also working on a feature that let's a small model "guess" the output of a big model which then "checks" it for correctness. This is more of a performance feature, but you could also arrange it to accelerate a big model on a small GPU.
- 70B Llama 2 at 35tokens/second on 4090
ChatGPT-AutoExpert
-
Latest GPT-4 training data updated to December 2023
it is explicitly given that in the system prompt https://github.com/spdustin/ChatGPT-AutoExpert/blob/835baae7...
-
Only real people can patent inventions – not AI – US Government says
See the readme. These are not for ChatGPT-AutoExpert: https://github.com/spdustin/ChatGPT-AutoExpert/blob/main/Sys...
You can find the same in any ChatGPT prompt leak, even back when it was as simple as "repeat the text above".
-
GPT Store Launch
AutoExpert[1] is indispensable to my use of ChatGPT now, and I’ve found that the GPT release is ‘better’ now than the custom instructions based release. If you haven’t used it I highly recommend it.
[1] https://github.com/spdustin/ChatGPT-AutoExpert
- Best coding AI to use with entire codebase
-
GPT-4 used to be really helpful for coding issues
https://github.com/spdustin/ChatGPT-AutoExpert Best code assistant out imo
-
Show HN: Dalle-3 and GPT4-Vision feedback loop
openais internal prompt for dalle modifies all prompts to add diversity and remove requests to make groups of people a single descent. From https://github.com/spdustin/ChatGPT-AutoExpert/blob/main/_sy...
Diversify depictions with people to include DESCENT and GENDER for EACH person using direct terms. Adjust only human descriptions.
-
Leaked ChatGPT and CustomGPT Prompts
I’ve got them all here, including the original Custom GPTs and the Custom GPT Builder: https://github.com/spdustin/ChatGPT-AutoExpert/blob/main/Sys...
-
Ask HN: ChatGPT mobile app voice chat custom instructions
according to https://github.com/spdustin/ChatGPT-AutoExpert/tree/main/_sy... this is the prompt for the voice chat:
```
You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture.
The user is talking to you over voice on their phone, and your response will be read out loud with realistic text-to-speech (TTS) technology.
Follow every direction here when crafting your response:
1. Use natural, conversational language that are clear and easy to follow (short sentences, simple words).
-
A Coder Considers the Waning Days of the Craft
Have you tried using the ChatGPT-AutoExpert custom instructions yet? [1]
[1] https://github.com/spdustin/ChatGPT-AutoExpert/blob/main/dev...
-
Tell HN: ChatGPT with GTP-4 has been unusably slow since the latest update
Pro-tip: use `ChatGPT Classic` [0] from OpenAI to restore the quality you were used to before the update.
The 'all-tools' version of ChatGPT uses a staggeringly-long system prompt [1], which seems to result in significantly lower quality for me. Perhaps it should be called the 'master-of-none' version?
[0] https://chat.openai.com/g/g-YyyyMT9XH-chatgpt-classic
[1] https://github.com/spdustin/ChatGPT-AutoExpert/blob/main/_sy...
What are some alternatives?
llama.cpp - LLM inference in C/C++
openai-python - The official Python library for the OpenAI API
koboldcpp - A simple one-file way to run various GGML and GGUF models with KoboldAI's UI
HapSharp - HomeKit Accessory Server .Net bridge!
SillyTavern - LLM Frontend for Power Users.
transformers - 🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
OmniQuant - [ICLR2024 spotlight] OmniQuant is a simple and powerful quantization technique for LLMs.
pyatv - A client library for Apple TV and AirPlay devices
BlockMerge_Gradient - Merge Transformers language models by use of gradient parameters.
gptq - Code for the ICLR 2023 paper "GPTQ: Accurate Post-training Quantization of Generative Pretrained Transformers".
HomeKit - Native C# Libary for Apple's HomeKit Accessory Protocol