Add modeling_stock_mamba.py

modeling_stock_mamba.py

@@ -1,7 +1,7 @@
 """Self-contained model definition shipped inside the Hugging Face repo.
 
 Depends only on torch, mamba_ssm (Mamba-3) and safetensors. Mirrors the training-time
-MambaBlock and MambaEncoder and the retired pretraining wrapper, so the exported
+MambaBlock and MambaEncoder and the pretraining wrapper, so the exported
 state_dict loads with no dependency on the training repo. A CUDA GPU is required: the
 Mamba-3 kernels have no CPU path and mamba_ssm fails at import on a CPU-only machine.
 """