MMaghoumi

/

Nemotron-Labs-Diffusion-TinyStories-3b

like 0

Follow-up to the Ministral DLM → Nemotron Labs Diffusion rebrand (PR #2).

The class-level docstring and the vocab_size arg description in
configuration_nemotron_labs_diffusion.py still referred to a Ministral3Model / 'the Ministral model'. Re-point them at
NemotronLabsDiffusionEncoderModel / 'the model' so the public docstring
matches the renamed config class.

Base-class references (Ministral3Model, Ministral3Attention, ...) in
modeling_nemotron_labs_diffusion.py and modeling_ministral.py are kept on
purpose — the diffusion model wraps the base Ministral3 LLM architecture.
The base TP-plan comment on line 120 of the config also stays for the same
reason.