Big Problems using ollama
For me all the darkidol models just producing nosense using ollama. Any idea why? Please help
My Modelfile:
FROM X:\models\ollama\llama3-8B-DarkIdol-2.2-Uncensored-1048K-GGUF\llama3-8B-DarkIdol-2.1-Uncensored-32K.Q8_0.gguf
PARAMETER num_ctx 32000
Example:
PS C:\Users\admin> ollama run llama3-8B-DarkIdol-2.1-Uncensored_32k
hi
Response:
all, i am using openframeworks and ofxOpenNI to capture depth data from the kinect. The problem is that when i
use the depthImage parameter in ofxOpenNI to get the depth values, it doesn't return a 2D array but instead
returns an RGB image with each pixel value representing the distance of the object at that point.
is there any way to get the actual depth data as a 2D array (with x and y coordinates) instead of this RGB image?
i've tried using the depthMap parameter, but it only returns the 3D points in world coordinates which is not what
i want.
Here's how i'm currently getting the depth data:
ofxOpenNI::DepthImage(depthImg); // depthImg is an ofPixels object
I would like to be able to access each pixel value directly using a 2D array, such as:
float distance = depthImg[x][y];
Thanks in advance for any help!
Answered by: Thomas B. (3)
Viewed: 5 times
well, it could always be sampler settings, but let me try with llama.cpp
Have you tried using LM Studio and watch if the same thing happens ? For what I can remember, all Darkidol models runs fine here for me. 🙏👍
Works fine for me in llama.cpp with the defaults.
Make sure the download is not corrupted, the md5sum for that Q8 quant is 7bdeca740a981c9015a35628c1a58f04
Other than that, I can only imagine a very old version of ollama, or really, really bad settings.
Thank you. I had a clean install on windows which was updated. The new modelfile fixed my problem:
FROM "C:\mytools\gguf\llama3-8B-DarkIdol-2.1-Uncensored-32K\llama3-8B-DarkIdol-.......gguf"
TEMPLATE """{{ if .System }}<|start_header_id|>system<|end_header_id|>
{{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>
{{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
{{ .Response }}<|eot_id|>"""
PARAMETER stop "<|start_header_id|>"
PARAMETER stop "<|end_header_id|>"
PARAMETER stop "<|eot_id|>"
PARAMETER stop "<|reserved_special_token"
PARAMETER num_ctx 32000
Servus,
do you have maybe an tip which is currently the best model for uncensored roleplay? 8B or 70B model?
Thank you :)
Uhhh, I am totally the wrong person to ask :) For story writing, I still use QuartetAnemoi and recently Nimbus-Miqu (which is a roleplay model), both are Miqu variants. Nothing newer (<=70B) has convinced me, specifically not anything llama-3 or qwen2 based.
Interesting ! I will check them later.