sentencepiece
KoboldAI
sentencepiece | KoboldAI | |
---|---|---|
19 | 41 | |
9,722 | 348 | |
2.5% | - | |
8.2 | 9.3 | |
10 days ago | 5 days ago | |
C++ | C++ | |
Apache License 2.0 | GNU Affero General Public License v3.0 |
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.
sentencepiece
- sentencepiece
-
LLM.int8(): 8-Bit Matrix Multiplication for Transformers at Scale
you need to train the model on 1 trillion tokens (https://platform.openai.com/tokenizer https://github.com/google/sentencepiece) anyways for it to get reasoning capacities, which it feels very unlikely that your data is that much.
I'm highly skeptical that you have enough data to pretrain if you don't have enough data to fine tune.
fine tuning + vector search + prompting of as much stuff as you can, on a LLM like palm2 or gpt4 is what I would do. otherwise you can use falcon 40B ofc.
maybe I should charge for this ahah
-
[P] TokenMonster Ungreedy ~ 35% faster inference and 35% increased context-length for large language models (compared to tiktoken). Benchmarks included.
a) Comparison with SentencePiece tokenizer with comparable settings (It can also ignore word-boundaries and create phrase tokens)
-
LLaMA tokenizer: is a JavaScript implementation available anywhere?
LLaMA uses the sentencepiece tokenizer: https://github.com/google/sentencepiece
-
[P] New tokenization method improves LLM performance & context-length by 25%+
Besides, are you familiar with SentencePiece? What you are doing looks very similar (generate a large vocab, prune worst token until vocab size is reached), only the token selection criterion is different. It's also purely data driven in the sense that there are no assumption specific to language (and it can optionally segment across whitespace, as you are doing).
-
Code runs without definition of function (automatically calls a different function instead)
Hi, I'm studying the implementation of encode and decode functions for Google's SentencePiece tokenizer.
-
How to handle multiple languages in a sentence?
I think many LMs nowadays use unicode tokenizers, that are not tied to specific languages. E.g. sentencepiece is the most popular one: https://github.com/google/sentencepiece
- Large language models are having their Stable Diffusion moment
-
LLaMA-7B in Pure C++ with full Apple Silicon support
If you are interested in implementing LLaMA yourself or learning, I noticed that the reference code by Facebook is one of the cleaner, easier to read ML code I've seen in a while. https://github.com/facebookresearch/llama/blob/main/llama/mo... It's about 200 lines long. You probably do need a bit of knowledge to understand what you are reading but I was pleasantly surprised.
For example in comparison, StableDiffusion torch code in diffusers and transformers Python libraries has lots of conditionals, experiments etc. that are not being used that can make it hard to follow what is going on.
Last weekend I got the "main loop" of the transformer working in pure CPU Rust code, following the reference code. My crappy code is just very very slow as I focused on getting it to run, not making it fast. The tokenizer uses some Google thing https://github.com/google/sentencepiece but luckily for inference it seems that you just need to be able to parse the tokenizer model file and not understand how it was created; I was able to strip out the protobuf files from that repository and add it to Rust and read the tokens.
I am optimistic that someone makes a high quality CPU or some CPU+GPU+SSD combination thingmaling that will make it somewhat practical to run even the large LLM models without needing an A100 or two.
- ChatGPT in an iOS Shortcut – Worlds Smartest HomeKit Voice Assistant
KoboldAI
-
LLM spews nonsense in CVE report for curl
It’s not that big a task as all that. There are a lot of unaligned models available, and user interfaces that aren’t that hard to use.
https://github.com/henk717/KoboldAI
-
Chat with, and help host, a free community LLM "horde"
https://github.com/henk717/KoboldAI
- Hosts pick a quantized community LLM to run, which is (IMO) the real magic of this system. Cloud services tend to run generic Llama chat/instruct models, OpenAI API models, or maybe a single proprietary finetune, but the Llama/Mistral finetuning community is red hot. New finetines and crazy merges/hybrids that outperform llama-chat in specific tasks (mostly Chat/Story/RP) come out every day, and each one has a different "flavor" and format:
https://huggingface.co/models?sort=modified&search=mistral+g...
- Run LLMs with KoboldaAI on Intel ARC
-
No idea what I'm doing help
Sourceforge is our official version but that one is to old to run newer models like Holomax, the releases for United can be found here : https://github.com/henk717/KoboldAI/releases
-
Still getting "read only" on JanitorAI even after setting model. Do I need to change anything config wise to get it to use pygmalion?
Colab Check: False, TPU: False INIT | OK | KAI Horde Models INFO | __main__::648 - We loaded the following model backends: KoboldAI API KoboldAI Old Colab Method Huggingface GooseAI Horde OpenAI Read Only INFO | __main__:general_startup:1363 - Running on Repo: https://github.com/henk717/koboldai Branch: INIT | Starting | Flask INIT | OK | Flask INIT | Starting | Webserver INIT | OK | Webserver MESSAGE | Webserver started! You may now connect with a browser at http://127.0.0.1:8501 INIT | Searching | GPU support INIT | Found | GPU support INIT | Starting | LUA bridge INIT | OK | LUA bridge INIT | Starting | LUA Scripts INIT | OK | LUA Scripts Setting Seed Traceback (most recent call last): File "B:\python\lib\site-packages\eventlet\hubs\selects.py", line 59, in wait listeners.get(fileno, hub.noop).cb(fileno) File "B:\python\lib\site-packages\eventlet\greenthread.py", line 221, in main result = function(*args, **kwargs) File "B:\python\lib\site-packages\eventlet\wsgi.py", line 837, in process_request proto.__init__(conn_state, self) File "B:\python\lib\site-packages\eventlet\wsgi.py", line 352, in __init__ self.finish() File "B:\python\lib\site-packages\eventlet\wsgi.py", line 751, in finish BaseHTTPServer.BaseHTTPRequestHandler.finish(self) File "B:\python\lib\socketserver.py", line 811, in finish self.wfile.close() File "B:\python\lib\socket.py", line 687, in write return self._sock.send(b) File "B:\python\lib\site-packages\eventlet\greenio\base.py", line 401, in send return self._send_loop(self.fd.send, data, flags) File "B:\python\lib\site-packages\eventlet\greenio\base.py", line 388, in _send_loop return send_method(data, *args) ConnectionAbortedError: [WinError 10053] An established connection was aborted by the software in your host machine Removing descriptor: 1488 Connection Attempt: 127.0.0.1 INFO | __main__:do_connect:2574 - Client connected! UI_1 TODO: Allow config INFO | modeling.inference_models.hf:set_input_parameters:189 - {'0_Layers': 18, 'CPU_Layers': 10, 'Disk_Layers': 0, 'class': 'model', 'label': 'PygmalionAI_pygmalion-6b', 'id': 'PygmalionAI_pygmalion-6b', 'name': 'PygmalionAI_pygmalion-6b', 'size': '', 'menu': 'Custom', 'path': 'C:\\KoboldAI\\models\\PygmalionAI_pygmalion-6b', 'ismenu': 'false', 'plugin': 'Huggingface'} INIT | Searching | GPU support INIT | Found | GPU support Loading checkpoint shards: 100%|█████████████████████████████████████████████████████████| 2/2 [00:19<00:00, 9.60s/it] Loading model tensors: 100%|##########| 56/56 [00:05<00:00, 9.52it/s]INIT | Starting | LUA bridge0, 8.93s/it] INIT | OK | LUA bridge INIT | Starting | LUA Scripts INIT | OK | LUA Scripts Setting Seed Connection Attempt: 127.0.0.1 INFO | __main__:do_connect:2574 - Client connected! UI_1
-
Kobold API URL for Chub Venus Ai
That is our developer version, its selectable in the Colab version dropdown and also available on https://github.com/henk717/koboldai
-
I got KoboldAI running on my computer and successfully connected it to Janitor, heres a small tutorial
Download Kobold from THIS LINK:https://github.com/henk717/KoboldAI. I downloaded Kobold from a different Github link and it wouldnt work, you need to get this specific one. Click on "code", then download zip
-
I created a repo on Github to categorize AI models. You can browse AIs from many categories!
https://github.com/henk717/KoboldAI https://github.com/LostRuins/koboldcpp/ https://github.com/ggerganov/llama.cpp https://github.com/AUTOMATIC1111/stable-diffusion-webui https://github.com/oobabooga/text-generation-webui
-
Meta’s new AI lets people make chatbots. They’re using it for sex.
For the third, I don't think Oobabooga supports the horde but KoboldAI does. I won't go into how to install KoboldAI since Oobabooga should give you enough freedom with 7B, 13B and maybe 30B models (depending on available RAM), but KoboldAI lets you download some models directly from the web interface, supports using online service providers to run the models for you, and supports the horde with a list of available models to choose from.
-
Kobold AI broke after update (New to this)
"Your Pytorch installation did not update correctly, you can solve this by running install_requirements.bat in the mode where it deletes the existing runtime. Alternative you can download a fresh copy of the offline installer for KoboldAI United from : https://github.com/henk717/KoboldAI/releases"
What are some alternatives?
transformers - 🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
text-generation-webui - A Gradio web UI for Large Language Models. Supports transformers, GPTQ, AWQ, EXL2, llama.cpp (GGUF), Llama models.
CTranslate2 - Fast inference engine for Transformer models
KoboldAI-Client
llama - Inference code for Llama models
koboldcpp - A simple one-file way to run various GGML and GGUF models with KoboldAI's UI
gpt-2 - Code for the paper "Language Models are Unsupervised Multitask Learners"
KoboldAI
stable-diffusion-webui - Stable Diffusion web UI
OpenNMT-Tutorial - Neural Machine Translation (NMT) tutorial. Data preprocessing, model training, evaluation, and deployment.
llama.cpp - LLM inference in C/C++