Git Product home page Git Product logo

Comments (8)

liar666 avatar liar666 commented on June 19, 2024

Before you ask, the rest of my config is pretty basic and has not changed:

  :config
  ;; Start server with ./llava-v1.5-7b-q4.llamafile --server --port 8000
  (gptel-make-openai "llama-cpp"          ;Any name
    :stream t                             ;Stream responses
    :protocol "http"
    :host "localhost:8000"                ;Llama.cpp server location
    :models '("test")
  )
  ;; OPTIONAL configuration
  (setq-default
        ;; HOME
        ;; gptel-model "gemma:2b"                        ; Pick your default model
        ;; WORK
        gptel-model "orionstar/orion14b-q4:latest"                        ; Pick your default model
        gptel-backend (gptel-make-ollama "gptel-localmodel" ; Any name of your choosing
                           :host "localhost:11434"          ; Where it's running
                           :stream t                        ; Stream responses
                           :models                          ; List of installed models
                           '(
                             ;; Code Specific Models
                             "codellama:latest"
                             "deepseek-coder:latest"
                             "stable-code:latest"
                             "starcoder:latest"
                             ;; Tiny Models
                             "dolphin-phi:latest"
                             "gemma:2b"
                             "orca-mini:latest"
                             "phi:latest"
                             "qwen:0"
                             "qwen:1"
                             "qwen:latest"
                             "stablelm-zephyr:latest"
                             "stablelm2:latest"
                             "tinydolphin:latest"
                             "tinyllama:latest"
                             ;; Medium Models
                             "gemma:7b"
                             "llama2-uncensored:latest"
                             "mistral:latest"
                             "vicuna:latest"
                             "zephyr:latest"
                             ;; Big Models
                             "mixtral:latest"
                             "llava:latest"
                             "dolphin-llama3:latest"
                             "orionstar/orion14b-q4:latest"
                             )
                      )
  )
  :bind
  (:map global-map
    ("C-S-g" . #'gptel-send)
  )
)

from gptel.

liar666 avatar liar666 commented on June 19, 2024

FYI, here's what I get in Ellama. Not perfect but way better:

** User:
generate the code of a tikz picture with 3 rectangles in blue 5 diamonds in red on their right and below a square of 3cm height with some lorem ipsum text in it


** Ellama:
Here' an example of a TikZ picture with three blue rectangles, five red diamonds, and a square below it filled with 'lorem ipsum' text: 
#+END_SRC
\begin{tikzpicture} 
    \foreach \i in {1,2,3}
        \draw[blue, rounded corners, draw=black] (\i*4-4,0) rectangle (+(-2,3));
    
    \foreach \i in {1,...,5}
        \draw[red]    (3.5+\i,1) node[octagon, minimum size=2cm] {};
    
    \draw[blue, square, very thick] (-4,-4) -- (-4,0) -- (0,0) -- (0,-4);
    
    \node at (-6.25,-2.5) {\textcolor{black}{\begin{paragraph}Lorem ipsum dolor sit amet,  \textbf{consectetur adipiscing elit}, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.\end{paragraph}}}; 
\end{tikzpicture}
#+END_SRC

from gptel.

karthink avatar karthink commented on June 19, 2024

What could be the problem?

If you are using gptel in Org mode, could you try text-mode or markdown-mode? I suspect it's an issue with the Org conversion that gptel does.

from gptel.

karthink avatar karthink commented on June 19, 2024

Also please try the following:

  1. (setq gptel-log-level 'info)
  2. Use gptel with Ollama
  3. Check the *gptel-log* buffer. You can paste its contents here.

from gptel.

liar666 avatar liar666 commented on June 19, 2024

What could be the problem?

If you are using gptel in Org mode, could you try text-mode or markdown-mode? I suspect it's an issue with the Org conversion that gptel does.

Result in fundamental-mode is exactly the same :(

Can you explain in formal but simple terms for a newbie what indoor bouldering is and how to proceed once you are in front of a wall?
Can you explain what the wall looks like and the principle with colored holds


 on an indoor climbing gym? What should be done in it?
Thank you. Please also provide some tips that beginner climbers are beginners? How do I want to use in your first time or what is involved in terms of how to practice at a beginner, what would you??
-. if when is and? with on an what
 the for??

? it can of it; of in. to and? you it
 ins: -?
 ?? and the is are to and? or is please
. , beginners? a beginner in in. that in be
? in .

?? what as your an. you in a? for use in in a have, what of what? b,. how it. and? when you b-. can on a an to are how. What if to with in?
 are the to help a what is an? to go we can understand? I will be available? When you could you'x? You is a you is it has been explained for example you needs like a get you give a explain in order you have some are able to use on your trywareveryncectional, so the basic The next when i? And what is looking a and make an other be more understanding is one-excessive b understand they say that I could be sure it can look them but you to climb if you know how is given in this means? is there you have fun some better of your said. Tokat
Example forks,
in the basics and the following to go on a are interested?s?

from gptel.

liar666 avatar liar666 commented on June 19, 2024

Also please try the following:

1. `(setq gptel-log-level 'info)`

2. Use gptel with Ollama

3. Check the `*gptel-log*` buffer.  You can paste its contents here.

Result is again BS:

Can you explain in formal but simple terms for a newbie what indoor bouldering is and how to proceed once you are in front of a wall?

Indoor bouldering is like?

Indoor Bouldering or rock climbing involves and what gear is needed and the process?
-the best way to understand it would be a beginner? I want to do the basic terms, please describe what the basics.

Begining of *gptel-log* ssems to show coorrect instruction are sent, but B is returned:

{
  "gptel": "request body",
  "timestamp": "2024-05-21 10:07:58"
}
{
  "model": "orionstar/orion14b-q4:latest",
  "messages": [
    {
      "role": "system",
      "content": "You are a large language model living in Emacs and a helpful assistant.  Respond concisely."
    },
    {
      "role": "user",
      "content": "Can you explain in formal but simple terms for a newbie what indoor bouldering is and how to proceed once you are in front of a wall?"
    }
  ],
  "stream": true,
  "options": {
    "temperature": 1.0
  }
}
{
  "gptel": "response body",
  "timestamp": "2024-05-21 10:08:11"
}
{
  "model": "orionstar/orion14b-q4:latest",
  "created_at": "2024-05-21T08:07:58.550080196Z",
  "message": {
    "role": "assistant",
    "content": "\n"
  },
  "done": false
}

Any idea what I could do wrong?

from gptel.

liar666 avatar liar666 commented on June 19, 2024

OK I'm so sorry... Apparently it's the orion model that outputs bullshit... I just tried it on its own:

$ ollama run orionstar/orion14b-q4:latest
>>> Can you explain in formal but simple terms for a newbie what indoor bouldering is and how to proceed once you are in front of a wall?
... 

Indoor bouldering means?

Sure! Indoor bouldering is like? What do I need to know more about this term "lead climbing wall? 😊

I naΓ―vely thought a larger model would be better... But finally using dolphin-llama3:latest give way better results!

from gptel.

liar666 avatar liar666 commented on June 19, 2024

OMG... And I'm really a dumbass... I had already configured ellama with dolphin-llama3:latest, thus the difference in answer quality...

from gptel.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    πŸ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. πŸ“ŠπŸ“ˆπŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❀️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.