Pomni

AI & ML interests

Pushing my GTX 1080 to its limits 🚀 I managed to get a 13B model with full GPU offload running once with relatively low quality loss.

Recent Activity

liked a model about 9 hours ago

openai/privacy-filter

updated a collection 1 day ago

Whisper Finetune GGML Quants

updated a collection 1 day ago

Whisper Finetune GGML Quants

View all activity

Organizations

None yet

liked a model about 9 hours ago

openai/privacy-filter

Token Classification • 1B • Updated 1 day ago • 1.89k • 584

updated a collection 1 day ago

Whisper Finetune GGML Quants

Collection

GGML quants of various Whisper finetunes, ranging from Q2_K all the way to F32. • 13 items • Updated 1 day ago

updated 2 models 2 days ago

Pomni/whisper-uz-ggml-allquants

Automatic Speech Recognition • Updated 2 days ago

Pomni/whisper-large-v3-turbo-german-ggml-allquants

Updated 2 days ago

published a model 2 days ago

Pomni/whisper-large-v3-turbo-german-ggml-allquants

Updated 2 days ago

published a model 3 days ago

Pomni/whisper-uz-ggml-allquants

Automatic Speech Recognition • Updated 2 days ago

reacted to Fourwheels2512's post with 🔥 13 days ago

Post

878

We trained Mistral 7B, Qwen 8B, Gemma 9B models on 5 domains sequentially to test catastrophic forgetting.
We achieved zero forgetting with medical knowledge retained at 100% after adding enterprise, finance, military, and real estate domains on top.
Most fine-tuned models catastrophically forget everything they learned when you train them on something new. We built a continual learning engine that prevents this. First of its kind.
We're shipping it as a SaaS platform at modelbrew.ai - dataset optimization + fine-tuning + continual learning in one pipeline.
I'm looking for ML fine-tuning engineers and researchers who want to test this. DM me or comment below.