@no_stp_on_snek: fine-tuning field notes how a model is built decides what you're even allowed to change. some models you can adjust fre…
Summary
A thread sharing field notes on fine-tuning, explaining how model architecture (e.g., heavily-quantized or mixture-of-experts) restricts which parts can be adjusted, urging practitioners to check model accessibility before planning work.
View Cached Full Text
Cached at: 07/04/26, 08:44 AM
fine-tuning field notes
how a model is built decides what you’re even allowed to change. some models you can adjust freely. others are so packed and compressed that most of the model is locked off-limits, so you can only touch a small accessible part and leave the rest frozen (heavily-quantized and mixture-of-experts models, mostly). same goal, but how far you can reach in depends entirely on the design. check what you can touch before you plan the work.
I plan to eventually
So, i wanted to test TwoTowers from Nvidia but it was a bit big. So i quantized it. but the speed is very slow. looks like vllm and llamacpp are still working on supporting block-diffusion.
So the hobbit is here… for whenever twotowers is supported in these inference engines.
https://huggingface.co/thetom-ai/Nemotron-TheHobbit-30B-A3B…
Similar Articles
@no_stp_on_snek: fine-tuning field notes you can give a model better judgment without teaching it anything new. i didn't add knowledge o…
The author shares field notes showing that fine-tuning can improve a model's judgment by steering attention without adding new knowledge or weights, effectively changing its instincts rather than its IQ.
@no_stp_on_snek: Fine Tuning Field Notes Article: short version: it was less about making it smarter and more about giving it a backbone…
A practitioner shares surprising findings from fine-tuning a small open model to be genuinely better in practical use, not just on benchmarks.
@no_stp_on_snek: fine-tuning field notes small behavior tweaks are a waterbed: push one spot down and another pops up. i trained a model…
A fine-tuning practitioner observes that adjusting one behavior in a model often causes unintended changes elsewhere, like a waterbed effect. Fixing pushback refusal silently broke strict formatting adherence, and subsequent fixes led to over-agreement or excessive refusal.
@no_stp_on_snek: what actually surprised me fine-tuning a small open model. note im failry new in this area so some of this may seem obv…
A developer shares surprising lessons from fine-tuning a small open model, including that base models often already max out on intended improvements, the real weakness is behavior (caving), and fine-tuning requires careful measurement and balancing.
@no_stp_on_snek: fine-tuning field notes every model family formats its conversations differently, and those formatting rules quietly wr…
A tweet warns that different AI model families have unique conversation formatting rules that silently corrupt training data, requiring developers to learn each family's quirks individually.