ANLP A3 cipher decryption: deltanet

Causal character tagger trained on data_new of ANLP Monsoon 2026 Assignment 3. Layer pattern DDDD (T = attention, M = Mamba, D = DeltaNet), model width 256, 3,453,467 parameters.

Validation character accuracy 0.1862, word accuracy 0.0153.

Load with src.part1.checkpoint.load_checkpoint from the assignment repository.

Downloads last month
30
Safetensors
Model size
3.45M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support