gemma4_text¶
Model type |
|
Class |
|
Task |
|
Source |
|
Description¶
Gemma 4 text-only causal language model.
Wraps :class:Gemma4TextModel (transformer backbone) with an lm_head
linear projection to produce next-token logits. Corresponds to HF
Gemma4ForCausalLM, which similarly wraps Gemma4Model (our
- class:
Gemma4TextModel) and adds the language-model head on top.
Registered as gemma4_text in the model registry. Uses hybrid
local/global attention, standard RMSNorm, and optional per-layer
input embeddings.
Usage¶
mobius build --model <MODEL_ID> output_dir/
from mobius import build
model = build("<MODEL_ID>")