impactframes / dolphin_llama3_omost

This is an Omost model, which has train on tags that define a canvas to improve the composition and prompt accuracy for stable diffusion generation using clip and T5 base encode systems. You can use this with https://github.com/if-ai/ComfyUI-IF_AI_tools

8B

417 Pulls Updated 2 months ago

Updated 2 months ago

2 months ago

3b8ea878d8dd · 4.9GB

parameters8.03B

quantizationQ4_K_M

{{ if .System }}<|im_start|>system {{ .System }}<|im_end|> {{ end }}{{ if .Prompt }}<|im_start|>user {{ .Prompt }}<|im_end|> {{ end }}<|im_start|>assistant {{ .Response }}<|im_end|>

Readme

The advantage of using this is that it seems to be at least twice as fast as other Omost implementations I tested and with better results.

Omost Guide

You can watch it here.

Note: For best results, it might help to set the temperature below 0.5.

A workflow is provided on the workflows folder of the repo - omost_agent_wf.json
You will need my ComfyUI custom node and select the Omost tool write an small prompt and run the queue Omost will take care of the rest

https://github.com/if-ai/ComfyUI-IF_AI_tools
and the ComfyUI Omost custom node for the rest of the conditioning nodes
https://github.com/huchenlei/ComfyUI_omost?tab=readme-ov-file

Prompt: “a jaguar on the rainy forest”
Model: ProteusV0.4
Non cherry picked but Omost picked
Results: