ERNIE-4.5-21B-A3B-PT | NVIDIA NeMo AutoModel

ERNIE-4.5-21B-A3B-PT has a checked-in NeMo AutoModel recipe for text generation. The Hugging Face configuration declares the Ernie4_5_MoeForCausalLM architecture.

Fine-Tune ERNIE-4.5-21B-A3B-PT

Follow the installation instructions, then run the recipe from the repository root:

automodel examples/llm_finetune/ernie4_5/ernie4_5_21b_a3b_hellaswag.yaml --nproc-per-node 8

Choose a Workflow

Goal	Start Here
Run the primary recipe for this model	Use the recipe configuration.
Prepare your environment	Follow the installation instructions.

Configuration

Setting	Configuration
Hardware	-
Strategy	FSDP2; tp_size=1, pp_size=1, cp_size=1, ep_size=8
Nodes	-
Features	Checkpointing (enabled=true), Activation checkpointing (activation_checkpointing=true), Transformer Engine, DeepEP (dispatcher=deepep)
Advanced	bfloat16, attention=te, linear=torch, experts=torch_mm, dispatcher=deepep

Model Reference

Model Architecture

Property	Value
Task	Text generation
Architecture	`Ernie4_5_MoeForCausalLM`

Available Models

baidu/ERNIE-4.5-21B-A3B-PT

LLM fine-tuning guide