You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

km_tokenizer_pertick_12x240

Grouped-FSQ keyboard/mouse tokenizer for PLAIOmni; each tick is encoded and decoded on its own. 8 raw 100 Hz control frames per block -> one 12.5 Hz action tick of 12 code ids over a per-group vocab of 240 (FSQ levels 8 x 6 x 5).

field	value
num_tokens	12 (2 key groups, 10 mouse groups)
fsq_levels / fsq_dim / vocab_size	(8, 6, 5) / 3 / 240
model_dim / ffn_dim / heads	256 / 1024 / 8
encoder / decoder / frame-encoder layers	6 / 6 / 2
mouse bins	96 linear + 24 geometric tail per axis (241 classes)
mouse bounds	dx +/-1280, dy +/-720
trained	epoch 12, step 150000

pytorch_model.bin is a fp32 state dict with the training module's model. prefix, loaded strictly by models.components.km_tokenizer_pertick.KmTokenizerPerTick.ensure_loaded.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support