Unanswered Questions
126 questions with no upvoted or accepted answers
6
votes
0
answers
536
views
I can't get my custom GPT to accept an API key - "error saving draft"
I'm trying to build a custom gpt with an action, basically the custom gpt can access an API if given proper instructions. I'm trying to get it to connect to plausible. I have built a simple code as a ...
- 161
5
votes
1
answer
255
views
How to fine-tune an LLM on a very small document
I am currently trying to fine-tune an LLM on a single document (about 1700 characters). I know that generally it is better to use prompt injection or something like a RAG system to provide specific ...
- 41
4
votes
0
answers
176
views
I'm writing a children's book and want to use Leonardo.ai to create all the characters in the same illustration style. Is that possible?
It seems that if I include the seed of the style I like, that's what creates consistency. Is that correct and complete? I'm a n00b, so if there's anything else you can suggest, I'd sure appreciate it....
- 41
3
votes
0
answers
105
views
How can I prompt an image generator to create scaled battle maps with a grid overlay for D&D or other table top games?
I have given instructions such as "grid overlay" "bird's eye view", yet the tool never seems to be able to scale properly. I am currently using Midjourneyon a PC with Windows 11.
- 3
3
votes
0
answers
141
views
How to avoid anti-aliasing in StableDiffusion output images?
I'm running stabilityai/stable-diffusion-xl-base-1.0 with the madebyollin/sdxl-vae-fp16-fix vae and custom lora weights.
The lora was trained on a dataset of thousands of 512x512 images that contained ...
3
votes
0
answers
50
views
What's the most universally compatible structure for storing audio recording datasets for voice cloning?
I would like to know the best way to store a voice actor's voice for reuse with multiple platforms.
Let's say I record my voice to be cloned, or pay a voice actor to narrate for the purpose of using ...
- 4,797
3
votes
0
answers
3k
views
Prompt for generating two characters in the same scene
There's a core issue with the AI tools that i'm having trouble with.
I generated my characters as I wanted, but then trying to combine them in a single image is very hard.
Character one:
https://cdn....
- 1
3
votes
0
answers
464
views
Poor concurrency and perf on multiple GPUs VM (running CodeLlama locally using Ollama)
I am trying to figure out if self hosting CodeLlama on a sufficiently powerful multi-GPU machine can be cost effective for my specific product needs.
However, when I go from 1 GPU to 4 GPU VM, I see ...
- 134
3
votes
0
answers
168
views
What are the temperature and top_p values used on chat.openai.com?
What are the temperature and top_p values used on https://chat.openai.com/? I know that the temperature and top_p values with the openai.com API both default to 1 for chat completion, but I don't know ...
3
votes
0
answers
127
views
Denoise image without converting to latent space?
KSampler requires using a variational autoencoder to convert an initial image to latent space before it tries to de-noise it. And this is awesome, but if you simply encode to latent space and decode ...
- 131
3
votes
0
answers
157
views
How to provide faces of public characters as a visual clue to Leonardo.ai?
I am a beginner in generative prompting and I am experimenting with Leonardo.ai. I like it's photoreal and cinematic effect, but it does not support upload of an image as input, apparently.
How can I ...
- 131
3
votes
0
answers
145
views
How is the "Corresponding requests per minute (RPM)" calculated when creating a GPT deployment in Azure?
I created an Azure OpenAI resource in my Azure account. I want to deploy a GPT-4-32k model.
How is the "Corresponding requests per minute (RPM)" calculated when creating a GPT deployment in ...
- 4,797
2
votes
0
answers
24
views
How can LLMs be prompted to avoid a wrap-up or coda?
LLMs in general seem to be very strongly constrained to provide responses that use the 3-act-play structure:
Set the context, or recap what came before.
Give the answer. <- This part is all that ...
- 121
2
votes
0
answers
34
views
Why is my LLM generating excessive code for a simple task?
I'm experimenting with LLMs on my local machine using Ollama. I've installed it and am running the CodeLlama 7b-code model:
ollama run codellama:7b-code
I gave it a simple prompt asking for a basic ...
2
votes
0
answers
76
views
How do people fine tune LLMs to not answer specific questions?
I am new to LLMs and learning new things daily.
I am right now trying to fine-tune Llama model that should avoid answering questions which are privacy related.
Is it a good practice to use model fine-...
- 21