PLaMo 3 NICT 2B Base

Model Description

PLaMo 3 NICT 2B Base is a 2B model pre-trained on English and Japanese datasets, developed by Preferred Networks, Inc. collaborative with National Institute of Information and Communications Technology, NICT.

PLaMo 3 NICT models adapt a hybrid architecture with Sliding Window Attention (SWA) and Traditional Attetntion layers.

PLaMo 3 NICT 2B Base is released under PLaMo community license. Please check the following license and agree to this before downloading.

• (EN) https://plamo.preferredai.jp/info/plamo-community-license-en
• (JA) https://plamo.preferredai.jp/info/plamo-community-license-ja

NOTE: This model has NOT been instruction-tuned for chat dialog or other downstream tasks.

For commercial users

Please check the PLaMo community license and contact us via the following form to use commercial purpose.

• (EN/JA) https://forms.gle/mTL8tBLrMYXKNZD56

Usage

Requirements

python>=3.10.0
numpy>=1.26.4
numba>=0.60.0
torch>=2.6.0,<=2.9.0
transformers>=4.56.0,<=4.57.1

Use vLLM

We provide the implementation via vLLM plugin.

See more details to setup the plugin for PLaMo 3 models: https://github.com/pfnet-research/vllm-plamo3-plugin

import vllm
model = vllm.LLM("pfnet/plamo-3-nict-2b-base", trust_remote_code=True)
response = model.generate(
    prompts=["The future of artificial intelligence technology is "],
    sampling_params=vllm.SamplingParams(
        n=1,
        max_tokens=32,
        top_k=50,
        top_p=0.95,
        temperature=1.0,
    ),
)[0]
generated_text = response.outputs[0].text
print(generated_text)

Load model directly

from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("pfnet/plamo-3-nict-2b-base", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("pfnet/plamo-3-nict-2b-base", trust_remote_code=True)
text = "これからの人工知能技術は"
input_ids = tokenizer(text, return_tensors="pt").input_ids
generated_tokens = model.generate(
    inputs=input_ids,
    max_new_tokens=32,
    do_sample=True,
    top_k=50,
    top_p=0.95,
    temperature=1.0,
)[0]
generated_text = tokenizer.decode(generated_tokens)
print(generated_text)

Model Details

• Model size: 2B
• Trained tokens: 200B
• Developed by:
  • Preferred Networks, Inc.
  • collaborative with National Institute of Information and Communications Technology, NICT
• Model type: Causal decoder-only
• Language(s): English, Japanese
• License: PLaMo community license

Training Dataset

We trained PLaMo 3 NICT 2B Base in two phases, phase 1 with 175B tokens and phase 2 with 25B tokens. The percentage of datasets in each phase is shown in the following table.

The training dataset also includes the following publicly available community resources:

• HuggingFaceFW/fineweb-2
• HuggingFaceTB/smollm-corpus
• bigcode/the-stack-v2-train-smol-ids
• nvidia/Nemotron-Pretraining-Dataset-sample
• EssentialAI/eai-taxonomy-stem-w-dclm

Tokenizer

PLaMo 3 NICT 2B Base tokenizer is optimized by numba, which is JIT compiler for numerical functions. The tokenizer is trained on a subst of the datasets for model pre-training.

Bias, Risks, and Limitations

PLaMo 3 NICT 2B Base is a new technology that carries risks with use. Testing conducted to date has been in English and Japanese, and has not covered, nor could it cover all scenarios. For these reasons, as with all LLMs, PLaMo 3 NICT 2B Base's potential outputs cannot be predicted in advance, and the model may in some instances produce inaccurate, biased or other objectionable responses to user prompts. Therefore, before deploying any applications of PLaMo 3 NICT 2B Base, developers should perform safety testing and tuning tailored to their specific applications of the model.

AI policies for Preferred Networks, Inc. group

• (EN) https://www.preferred.jp/en/company/aipolicy/
• (JA) https://www.preferred.jp/ja/company/aipolicy/