Module neural_model 

NeuralModel – bit-level cross-context model.

Uses context hashes that combine bit-level information from the current byte being decoded with byte-level context. This captures patterns that traditional byte-context CM models miss because they don’t condition on the bits already decoded in the current byte.

Key contexts:

• c0_full (all decoded bits so far, not just low 8) x c1 nibbles
• Byte boundary contexts (position in line, word boundary)
• Bit-pattern contexts (repeated bit runs, alternating patterns)

Uses the same ContextModel (ContextMap + StateMap) machinery as other models.

CRITICAL: Encoder and decoder must produce IDENTICAL neural model state.

Structs

NeuralModel

Neural model using bit-level cross-contexts.