Useful resources on the Internet for working with Large Language Models.

There are many useful resources on the Internet for working with Large Language resources. Such as comparing Large Language Models side by side. Visit this website. This allows comparing two random Chatbots side by side and then ranking the performance of each one.

This is the site. Give the chatbots a request and see the performance of each one. Given a programming challenge, it is striking how each one handles it.

https://chat.lmsys.org/.

It is also possible to run a model locally on your PC. This requires a lot of processing power and GPU VRAM. So a 24 GiB card would be better than a card with 6GiB. But the process of training a model would be the fun part. Find a model here to download and train. Runs locally on YOUR CPU, on YOUR model you trained yourself (but there are some available for download), and it’s Open Source.

https://github.com/nomic-ai/gpt4all.

Other promising projects or those that have future potential: https://github.com/Stability-AI/StableLM/ (demo in https://stablediffusionweb.com/StableLM ).

https://github.com/tatsu-lab/stanford_alpaca https://github.com/tloen/alpaca-lora (from what I understand it is like an optimized version of the previous one to use on modest hardware).

https://github.com/mosaicml/llm-foundry/ (more specifically https://huggingface.co/mosaicml/mpt-7b-chat but other models are available at https://www.mosaicml.com/blog/mpt-7b ).

https://huggingface.co/BlinkDL/rwkv-4-raven and https://huggingface.co/BlinkDL/rwkv-4-world https://github.com/LAION-AI/Open-Assistant (there is a demo at https://open-assistant.io/ but it requires registration).

https://huggingface.co/lmsys/fastchat-t5-3b-v1.0 https://github.com/nlpxucan/WizardLM https://github.com/artidoro/qlora (demo in https://huggingface.co/spaces/uwnlp/guanaco-playground-tgi ).

https://github.com/young-geng/EasyLM/blob/main/docs/koala.md.

https://github.com/databrickslabs/dolly.

https://github.com/THUDM/ChatGLM-6B/blob/main/README_en.md.

In addition to the demos mentioned, you can also try all previous models at https://chat.lmsys.org except rwkv-4-world and StableLM (by default it loads vicuna-13b).

Stanford Alpaca Large Language Model.

https://github.com/tatsu-lab/stanford_alpaca.

These resources should be very useful for setting up your own GPT-4 chatbot. Until the Google LLM is released, this is supposed to be far more massive than GPT-4. PaLM 2 is going to be good competitor to any other Large Language Model. Try Google Bard in the meantime.

https://bard.google.com/chat/.

I asked Google Bard to make this request.

Write a simple program in C for Linux to read all files from a directory and print the output.

And I got this response.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

#include <dirent.h>
#include <stdio.h>
 
int main() {
  DIR *dir;
  struct dirent *entry;
 
  // Open the directory.
  dir = opendir(".");
  if (dir == NULL) {
    perror("opendir");
    return 1;
  }
 
  // Read all entries in the directory.
  while ((entry = readdir(dir)) != NULL) {
    // Check if the entry is a regular file.
    if (entry->d_type == DT_REG) {
      // Print the filename.
      printf("%s\n", entry->d_name);
    }
  }
 
  // Close the directory.
  closedir(dir);
 
  return 0;
}

This program listed the directory instead, but it did work right away with no fuss.