TENSORLIGHT

Will it fit?

Paste any model from the Hub. Its config says how much memory it needs and how many bytes each token reads; the chip’s spec sheet says the rest.

Weights
KV cache