MiniCPM5-1B

View as Markdown

MiniCPM is a compact language model series from OpenBMB and Tsinghua University, designed to deliver strong performance at small parameter counts using model merging and continuous training techniques.

Set up NeMo AutoModel with the latest container or follow the installation instructions.

Fine-Tune MiniCPM5-1B

From the repository root, run:

# Full fine-tuning
uv run automodel examples/llm_finetune/minicpm/minicpm5_1b_squad.yaml --nproc-per-node 8
# LoRA fine-tuning
uv run automodel examples/llm_finetune/minicpm/minicpm5_1b_squad_peft.yaml --nproc-per-node 8

Model Reference

Model Architecture

PropertyValue
TaskText Generation
ArchitectureMiniCPMForCausalLM / MiniCPM3ForCausalLM / MiniCPM5ForCausalLM
Parameters1B
Hugging Face Organizationopenbmb
  • MiniCPMForCausalLM - MiniCPM v1/v2
  • MiniCPM3ForCausalLM - MiniCPM3
  • MiniCPM5ForCausalLM - MiniCPM5

Available Models

ModelHF ID
MiniCPM5 1Bopenbmb/MiniCPM5-1B