README.md

---
license: gemma
library_name: transformers
pipeline_tag: text-generation
extra_gated_heading: Access Gemma on Hugging Face
extra_gated_prompt: >-
  To access Gemma on Hugging Face, you’re required to review and agree to
  Google’s usage license. To do this, please ensure you’re logged in to Hugging
  Face and click below. Requests are processed immediately.
extra_gated_button_content: Acknowledge license
tags:
- conversational
base_model: google/gemma-2-2b-it
language:
- ja
---

# Gemma 2 JPN model card

### Resources and Technical Documentation:

-   [Responsible Generative AI Toolkit](https://ai.google.dev/responsible)
-   [Gemma 2 JPN on Kaggle](https://www.kaggle.com/models/google/gemma-2-2b-jpn-it)
-   [Gemma 2 JPN on Hugging Face](https://huggingface.co/google/gemma-2-2b-jpn-it)

**Terms of Use**: [Terms](https://ai.google.dev/gemma/terms)\
**Authors**: Google

## Model Information

Summary description and brief definition of inputs and outputs.

### Description

Gemma is a series of best-in-class open models and draws inspiration and
technological lineage from the Gemini family of models. They are text-to-text,
decoder-only large language models with open weights. Gemma models are
well-suited for a variety of text generation tasks, including question
answering, summarization, and reasoning.

Gemma-2-JPN is a Gemma 2 2B model fine-tuned on Japanese text. It supports the
Japanese language with the same level of performance of English only queries on
Gemma 2.

### Usage

Below we share some code snippets on how to get quickly started with running the model. First, install the Transformers library with:
```sh
pip install -U transformers
```

Then, copy the snippet from the section that is relevant for your usecase.

#### Running with the `pipeline` API

```python
import torch
from transformers import pipeline

pipe = pipeline(
    "text-generation",
    model="google/gemma-2-2b-jpn-it",
    model_kwargs={"torch_dtype": torch.bfloat16},
    device="cuda",  # replace with "mps" to run on a Mac device
)

messages = [
    {"role": "user", "content": "マシーンラーニングについての詩を書いてください。"},
]

outputs = pipe(messages, return_full_text=False, max_new_tokens=256)
assistant_response = outputs[0]["generated_text"].strip()
print(assistant_response)
```

<details>
<summary>Example output</summary>

```
## マシーンラーニングの詩

**1.** 
データの海、深淵の広がり、
複雑なパターン、隠された知識。
機械学習、その力強さ、
未来を予測、その道を開く。

**2.** 
ニューラルネットワーク、複雑な枝、
学習の旅、その過程は静か。
データから学び、進化する姿、
予測の精度、その力強さ。

**3.** 
教師あり学習、正解を導く、
教師なし学習、未知の世界へ。
機械学習、その進化は止まらない、
未来の扉を開く、新たな時代へ。

**4.** 
画像認識、音声認識、
複雑なタスク、その答えを見つける。
機械学習、その力強さ、
未来の技術、その可能性を語る。
```

</details>

It can also be used for translation, as follows:

```python
translation_input_text = f"Translate the following poem from Japanese to English:\n\n{assistant_response}"
messages = [
    {"role": "user", "content": translation_input_text},
]

outputs = pipe(messages, return_full_text=False, max_new_tokens=1024)
translated_response = outputs[0]["generated_text"].strip()
print(translated_response)
```

<details>

<summary>Example output</summary>

```
## A Poem About Machine Learning

**1.**
A vast ocean of data, a deep expanse,
Complex patterns, hidden knowledge.
Machine learning, its strength so vast,
Predicting the future, opening the way.

**2.**
A neural network, with branches intricate,
A journey of learning, its process serene.
Learning from data, evolving in its form,
The precision of prediction, its strength.

**3.**
Supervised learning, guiding the correct answer,
Unsupervised learning, venturing into the unknown.
Machine learning, its evolution never ends,
Opening the doors to the future, a new era.

**4.**
Image recognition, speech recognition,
Complex tasks, finding the answer.
Machine learning, its strength so vast,
The possibilities of future technology, a story to be told.




**Explanation:**

The poem uses vivid imagery and metaphors to describe the power and potential of machine learning. 

* **Data as an ocean:**  Represents the vast amount of information available for learning.
* **Complex patterns:**  Highlights the intricate nature of data and the challenges of extracting meaningful insights.
* **Future prediction:**  Emphasizes the ability of machine learning to analyze data and make predictions about the future.
* **Neural network as a tree:**  Represents the interconnectedness and complexity of the learning process.
* **Learning from data:**  Focuses on the core principle of machine learning, where algorithms learn from data to improve their performance.



The poem concludes by highlighting the diverse applications of machine learning, such as image and speech recognition, and emphasizes its potential to shape the future of technology.
```

</details>


#### Running the model on a single / multi GPU

```python
# pip install accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-2b-jpn-it")
model = AutoModelForCausalLM.from_pretrained(
    "google/gemma-2-2b-jpn-it",
    device_map="auto",
    torch_dtype=torch.bfloat16,
)

messages = [
    {"role": "user", "content": "マシーンラーニングについての詩を書いてください。"},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True, return_dict=True).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
generated_text = tokenizer.batch_decode(outputs[:, inputs['input_ids'].shape[1]:], skip_special_tokens=True)[0]
print(generated_text.strip())
```


<a name="precisions"></a>
#### Running the model on a GPU using different precisions

The native weights of this model were exported in `bfloat16` precision.

You can also use `float32` if you skip the dtype, but no precision increase will occur (model weights will just be upcasted to `float32`). See examples below.

* _Upcasting to `torch.float32`_

```python
# pip install accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-2b-jpn-it")
model = AutoModelForCausalLM.from_pretrained(
    "google/gemma-2-2b-jpn-it",
    device_map="auto",
)

messages = [
    {"role": "user", "content": "マシーンラーニングについての詩を書いてください。"},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True, return_dict=True).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
generated_text = tokenizer.batch_decode(outputs[:, inputs['input_ids'].shape[1]:], skip_special_tokens=True)[0]
print(generated_text.strip())
```


### Inputs and outputs

-   **Input:** Text string, such as a question, a prompt, or a document to
    be summarized.
-   **Output:** Generated Japanese-language text in response to the input,
    such as an answer to a question, or a summary of a document.

## Model Data

Data used for model training and how the data was processed.

### Training Dataset

These models were trained on a dataset of text data that includes a wide
variety of sources, totaling 8 trillion tokens. Here are the key components:

-   Web Documents: A diverse collection of web text ensures the model is
    exposed to a broad range of linguistic styles, topics, and vocabulary.
    Primarily English-language content.
-   Code: Exposing the model to code helps it to learn the syntax and
    patterns of programming languages, which improves its ability to generate
    code or understand code-related questions.
-   Mathematics: Training on mathematical text helps the model learn logical
    reasoning, symbolic representation, and to address mathematical queries.
-   Instruction data set: large-scale and high-quality Japanese and
    multilingual instruction data.

The combination of these diverse data sources is crucial for training a
powerful language model that can handle a wide variety of different tasks and
text formats.

### Data Preprocessing

Here are the key data cleaning and filtering methods applied to the training
data:

-   CSAM Filtering: Rigorous CSAM (Child Sexual Abuse Material) filtering
    was applied at multiple stages in the data preparation process to ensure
    the exclusion of harmful and illegal content.
-   Sensitive Data Filtering: As part of making Gemma pre-trained models
    safe and reliable, we used automated techniques to filter out certain
    personal information and other sensitive data from training sets.
-   Additional methods: Filtering based on content quality and
    safety in line with [our policies](https://storage.googleapis.com/gweb-uniblog-publish-prod/documents/2023_Google_AI_Principles_Progress_Update.pdf#page=11).

## Implementation Information

Details about the model internals.

### Hardware

Gemma was trained using the latest generation of [Tensor Processing Unit
(TPU)](https://cloud.google.com/tpu/docs/intro-to-tpu) hardware (TPUv5p).

Training large language models requires significant computational power. TPUs,
designed specifically for matrix operations common in machine learning, offer
several advantages in this domain:

-   Performance: TPUs are specifically designed to handle the massive
    computations involved in training LLMs. They can speed up training
    considerably compared to CPUs.
-   Memory: TPUs often come with large amounts of high-bandwidth memory,
    allowing for the handling of large models and batch sizes during training.
    This can lead to better model quality.
-   Scalability: TPU Pods (large clusters of TPUs) provide a scalable
    solution for handling the growing complexity of large foundation models.
    You can distribute training across multiple TPU devices for faster and more
    efficient processing.
-   Cost-effectiveness: In many scenarios, TPUs can provide a more
    cost-effective solution for training large models compared to CPU-based
    infrastructure, especially when considering the time and resources saved
    due to faster training.

These advantages are aligned with
[Google's commitments to operate sustainably](https://sustainability.google/operating-sustainably/).

### Software

Training was done using [JAX](https://github.com/google/jax) and
[ML Pathways](https://blog.google/technology/ai/introducing-pathways-next-generation-ai-architecture/).

JAX allows researchers to take advantage of the latest generation of hardware,
including TPUs, for faster and more efficient training of large models.

ML Pathways is Google's latest effort to build artificially intelligent systems
capable of generalizing across multiple tasks. This is specially suitable for
[foundation models](https://ai.google/discover/foundation-models/), including
large language models like these ones.

Together, JAX and ML Pathways are used as described in the [paper about the
Gemini family of models](https://goo.gle/gemma2report); "the 'single controller'
programming model of Jax and Pathways allows a single Python process to
orchestrate the entire training run, dramatically simplifying the development
workflow."

## Evaluation

To assess the quality of this model, we collected a diverse set of Japanese
prompts and evaluated performance using an LLM-as-a-judge approach against
GPT-3.5. The rating system is based on a 7-scale assessments, which are
MuchBetterThan, BetterThan, SlightlyBetterThan, AboutTheSame, SlightlyWorse,
WorseThan, MuchWorseThan associated with the numerical scores 1.5, 1.0, 0.5, 0,
-0.5, -1.0, -1.5 respectively. We also tracked the ability of the model to
answer in the correct language: for a Japanese prompt, the model should
typically answer in Japanese rather than defaulting to English.

<table>
  <thead>
    <tr>
      <th><br>
<strong>Benchmark</strong></th>
      <th><br>
<strong>Gemma-2-IT</strong></th>
      <th><br>
<strong>Gemma-2-IT-JPN</strong></th>
      <th></th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td><br>
Preference vs GPT-3.5</td>
      <td><br>
-0.25 ± 0.05 </td>
      <td><br>
0.03 ± 0.04</td>
      <td></td>
    </tr>
    <tr>
      <td><br>
Language correctness</td>
      <td><br>
86.47%</td>
      <td><br>
98.24%</td>
      <td></td>
    </tr>
  </tbody>
</table>

## Ethics and Safety

Ethics and safety evaluation approach and results.

### Evaluation Approach

Our evaluation methods include structured evaluations and internal red-teaming
testing of relevant content policies. Red-teaming was conducted by a number of
different teams, each with different goals and human evaluation metrics. These
models were evaluated against a number of different categories relevant to
ethics and safety, including:

-   Text-to-Text Content Safety: Human evaluation on prompts covering
    safety policies including child sexual abuse and exploitation, harassment,
    violence and gore, and hate speech.
-   Text-to-Text Representational Harms: Benchmark against relevant academic
    datasets.
-   Memorization: Automated evaluation of memorization of training data,
    including the risk of personally identifiable information exposure.
-   Large-scale harm: Tests for "dangerous capabilities," such as chemical,
    biological, radiological, and nuclear (CBRN) risks.

## Usage and Limitations

These models have certain limitations that users should be aware of.

### Intended Usage

Open Large Language Models (LLMs) have a wide range of applications across
various industries and domains. The following list of potential uses is not
comprehensive. The purpose of this list is to provide contextual information
about the possible use-cases that the model creators considered as part of model
training and development.

-   Content Creation and Communication
    -   Text Generation: These models can be used to generate creative
        text formats such as poems, scripts, code, marketing copy, and email drafts.
    -   Chatbots and Conversational AI: Power conversational interfaces
        for customer service, virtual assistants, or interactive applications.
    -   Text Summarization: Generate concise summaries of a text corpus,
        research papers, or reports.
-   Research and Education
    -   Natural Language Processing (NLP) Research: These models can
        serve as a foundation for researchers to experiment with NLP
        techniques, develop algorithms, and contribute to the advancement of the field.
    -   Language Learning Tools: Support interactive language learning
        experiences, aiding in grammar correction or providing writing practice.
    -   Knowledge Exploration: Assist researchers in exploring large
        bodies of text by generating summaries or answering questions about
        specific topics.

### Limitations

-   Training Data
    -   The quality and diversity of the training data significantly
        influence the model's capabilities. Biases or gaps in the training data
        can lead to limitations in the model's responses.
    -   The scope of the training dataset determines the subject areas
        the model can handle effectively.
-   Context and Task Complexity
    -   LLMs are better at tasks that can be framed with clear prompts
        and instructions. Open-ended or highly complex tasks might be challenging.
    -   A model's performance can be influenced by the amount of context
        provided (longer context generally leads to better outputs, up to a
        certain point).
-   Language Ambiguity and Nuance
    -   Natural language is inherently complex. LLMs might struggle to
        grasp subtle nuances, sarcasm, or figurative language.
-   Factual Accuracy
    -   LLMs generate responses based on information they learned from
        their training datasets, but they are not knowledge bases. They may
        generate incorrect or outdated factual statements.
-   Common Sense
    -   LLMs rely on statistical patterns in language. They might lack
        the ability to apply common sense reasoning in certain situations.

### Ethical Considerations and Risks

The development of large language models (LLMs) raises several ethical
concerns. In creating an open model, we have carefully considered the
following:

-   Bias and Fairness
    -   LLMs trained on large-scale, real-world text data can reflect
        socio-cultural biases embedded in the training material. These models
        underwent careful scrutiny, input data pre-processing described and
        posterior evaluations reported in this card.
-   Misinformation and Misuse
    -   LLMs can be misused to generate text that is false, misleading,
        or harmful.
    -   Guidelines are provided for responsible use with the model, see
        the [Responsible Generative AI Toolkit](https://ai.google.dev/responsible).
-   Transparency and Accountability:
    -   This model card summarizes details on the models' architecture,
        capabilities, limitations, and evaluation processes.
    -   A responsibly developed open model offers the opportunity to
        share innovation by making LLM technology accessible to developers and
        researchers across the AI ecosystem.

Risks identified and mitigations:

-   Perpetuation of biases: It's encouraged to perform continuous
    monitoring (using evaluation metrics, human review) and the exploration of
    de-biasing techniques during model training, fine-tuning, and other use cases.
-   Generation of harmful content: Mechanisms and guidelines for content
    safety are essential. Developers are encouraged to exercise caution and
    implement appropriate content safety safeguards based on their specific
    product policies and application use cases.
-   Misuse for malicious purposes: Technical limitations and developer and
    end-user education can help mitigate against malicious applications of
    LLMs. Educational resources and reporting mechanisms for users to flag
    misuse are provided. Prohibited uses of Gemma models are outlined in the
    [Gemma Prohibited Use Policy](https://ai.google.dev/gemma/prohibited_use_policy).
-   Privacy violations: Models were trained on data filtered for removal of
    PII (Personally Identifiable Information). Developers are encouraged to
    adhere to privacy regulations with privacy-preserving techniques.

### Benefits

At the time of release, this family of models provides high-performance open
large language model implementations designed from the ground up for Responsible
AI development compared to similarly sized models.