mamba¶
Model type |
|
Class |
|
Task |
|
Source |
|
Description¶
Mamba causal language model with SSM backbone and LM head.
Uses selective state space layers instead of transformer attention. Carries conv_state + ssm_state per layer (no KV cache).
Inputs: input_ids, past_states (conv + ssm per layer). Outputs: logits, present_states.
HuggingFace reference: MambaForCausalLM.
Usage¶
mobius build --model <MODEL_ID> output_dir/
from mobius import build
model = build("<MODEL_ID>")