I'm pretty new to all of this ml stuff, so I wonder, I had models on colab that I redownloaded from hf with the auto installl, my older models seem to have a numeric annotation behind them that makes them different in title. I thought at first this was due to some way that colab zipped them, but now I'm wondering if that is a significator of some sort to the number of epochs and training they may have seen through use, and if so am I losing my "WORK" from previous training by rolling back to the HF originals? Maybe I am overthinking it, but is there a good introductory explanation of this and the process of further training of ckpt / safetensor files?
#Losing your work when loading models?
terse gazelle
tawdry reef
Your models are static, and aren't being "trained" through using them to generate images. If you are doing actual model training (e.g., fine-tuning and training via textual inversion, dreambooth, etc.), those would be generating completely new files.
terse gazelle
Any idea why they are being truncated with numericals then?
tawdry reef
Example?
tawdry reef
yeah, not clear, but i'm 99% sure that those were named that on download.
the numbers mean nothing, and affect nothing.
you could rename the model whatever you want - just has to have that name in the call out in the models.yaml file
terse gazelle
How do I remove models I added using !invoke ? I dont see them in the yaml to delete.
I have one or two that dont load for some reason
tawdry reef
You are likely looking at the wrong yaml file