Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
5 changes: 5 additions & 0 deletions ja/blog.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
title: W&B チュートリアル & ブログ
url: https://wandb.ai/fully-connected
---

5 changes: 5 additions & 0 deletions ja/courses.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
---
title: W&B コース
url: https://wandb.ai/site/courses/?e-filter-32fa548-course-category=wb-platform
---

78 changes: 78 additions & 0 deletions ja/get-started.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,78 @@
---
title: Weights & Biases を使ってみる
description: ユースケース に適した W&B 製品を選択し、 開始方法 を確認しましょう
---

## 製品の比較

Weights & Biases へようこそ! 製品を使い始める前に、ユースケースに最適な製品を確認することが重要です。

| 製品 | 最適な用途 | 主な機能 |
|---------|----------|--------------|
| **[W&B Models](#w%26b-models)** | ML モデルをゼロからトレーニングする | 実験管理、ハイパーパラメーター最適化、モデルレジストリ、可視化 |
| **[W&B Weave](#w%26b-weave)** | LLM アプリケーションを構築する | トレース、プロンプト管理、評価、プロダクション AI アプリのコスト追跡 |
| **[W&B Inference](#w%26b-inference)** | 学習済みモデルを使用する | ホストされたオープンソースモデル、API アクセス、テスト用のモデルプレイグラウンド |
| **[W&B Training](#w%26b-training)** | モデルをファインチューニングする | 強化学習を用いた LoRA やカスタムモデル適応の作成とデプロイ |

## W&B Models

<Columns cols={2}>
<Card title="Models クイックスタート" href="/models/quickstart">
W&B の「hello world」であり、最初のデータをログに記録する方法を説明します。
</Card>
<Card title="Models を始める" href="/models/models_quickstart">
実際の ML 実験を使用して、Models 製品全体を体験する本格的なチュートリアルです。
</Card>
<Card title="W&B 101 コース" href="https://wandb.ai/site/courses/101/">
実験管理に重点を置いたビデオ形式のコースです。理解度を確認するためのクイズも用意されています。
</Card>
<Card title="YouTube チュートリアル" href="https://www.youtube.com/watch?v=tHAFujRhZLA">
モデルのトレーニング、評価、開発、デプロイの方法と、ライフサイクルの各ステップで wandb を活用して、より高性能なモデルをより速く構築する方法を学びます。
</Card>
</Columns>


## W&B Weave

<Columns cols={2}>
<Card title="Weave クイックスタート" href="/weave/quickstart">
コードをデコレートして LLM を呼び出すことで、Weave トレースを記録し、完璧な LLM ワークフローへの道を歩み始める方法を学びます。
</Card>
<Card title="W&B Inference で Weave を学ぶ" href="/weave/quickstart-inference">
W&B Inference でホストされている様々なモデルのパフォーマンスを Weave で実際に評価する、本格的なチュートリアルです。
</Card>
<Card title="W&B Weave 101 コース" href="https://site.wandb.ai/courses/weave/">
言語モデルのワークフローをログに記録し、デバッグし、評価する方法を学ぶビデオ形式のコースです。理解度を確認するためのクイズも用意されています。
</Card>
<Card title="YouTube デモ" href="https://www.youtube.com/watch?v=IQcGGNLN3zo">
AI アプリケーションを継続的に評価、監視、反復し、品質、レイテンシ、コスト、安全性を向上させる方法を学びます。
</Card>
</Columns>

## W&B Inference

<Columns cols={2}>
<Card title="Inference イントロダクション" href="/inference">
標準的な OpenAI REST API を使用して、W&B Inference でホストされている任意のモデルを呼び出す方法を示すクイックスタートが含まれています。
</Card>
<Card title="W&B Inference で Weave を学ぶ" href="/weave/quickstart-inference">
W&B Inference でホストされている様々なモデルのパフォーマンスを Weave で実際に評価する、本格的なチュートリアルです。
</Card>
<Card title="Inference プレイグラウンドを試す" href="https://wandb.ai/inference">
W&B Inference は非常に簡単に使用できます。ホストされているモデルをクリックし、プロンプトを試し、可観測性レイヤーが動作する様子を確認してください。
</Card>
<Card title="例" href="/inference/examples">
W&B Inference が一般的な LLM への呼び出しをトレースし、結果を評価するいくつかの簡単な例を実行します。
</Card>
</Columns>

## W&B Training

<Columns cols={2}>
<Card title="クイックスタート" href="https://art.openpipe.ai/getting-started/quick-start">
OpenPipe の ART ライブラリと共に W&B Training を使用して、2048 ゲームをプレイするモデルをトレーニングします。
</Card>
<Card title="トレーニング済みモデルを使用する" href="/training/serverless-rl/use-trained-models">
トレーニング済みモデルを作成した後、それをコード内で使用する方法を学びます。
</Card>
</Columns>
67 changes: 67 additions & 0 deletions ja/index.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,67 @@
---
title: Weights & Biases ドキュメンテーション
description: Weights & Biases のすべての製品のドキュメントを表示する
sidebarTitle: Home
mode: wide
---
import {Banner} from "/snippets/Banner.jsx";
import {HomeWrapper} from "/snippets/home.jsx";
import {ProductCard} from "/snippets/ProductCard.jsx";

<HomeWrapper>
<div className="grid grid-cols-1 md:grid-cols-2 gap-4">
<ProductCard
title="W&B Models"
iconSrc="icons/cropped-models.svg"
href="/models"
subtitle="AI モデルの開発"
>
W&B Models を使用して、 AI モデルの開発を管理します。トレーニング、ファインチューニング、 Reports 、ハイパーパラメーター Sweeps の自動化、バージョン管理と再現性のための Model Registry などの機能が含まれます。
<br/>
<br/>
• <a href="/models">イントロダクション</a><br/>
• <a href="/models/quickstart">クイックスタート</a><br/>
• <a href="https://www.youtube.com/watch?v=tHAFujRhZLA">YouTube チュートリアル</a><br/>
</ProductCard>

<ProductCard
title="W&B Weave"
iconSrc="icons/cropped-weave.svg"
href="/weave"
subtitle="アプリでの AI モデルの活用"
>
W&B Weave を使用して、コード内の AI モデルを管理します。トレース、出力の評価、コスト見積もり、ホストされた推論サービス、および異なる大規模言語モデル( LLM )や設定を比較するためのプレイグラウンドなどの機能が含まれます。
<br/>
<br/>
• <a href="/weave">イントロダクション</a><br/>
• <a href="/weave/quickstart">クイックスタート</a><br/>
• <a href="https://www.youtube.com/watch?v=IQcGGNLN3zo">YouTube デモ</a><br/>
</ProductCard>

<ProductCard
title="W&B Inference"
iconSrc="icons/cropped-inference.svg"
href="/inference"
subtitle="基盤モデルへのアクセス"
>
W&B Inference を使用して、 OpenAI 互換の API を通じて主要なオープンソース基盤モデルにアクセスします。複数のモデルオプション、使用状況のトラッキング、トレースや評価のための Weave とのインテグレーションなどの機能が含まれます。
<br/>
<br/>
• <a href="/inference">イントロダクション</a><br/>
• <a href="https://wandb.ai/inference">プレイグラウンドで試す</a>
</ProductCard>

<ProductCard
title="W&B Training"
iconSrc="icons/cropped-training.svg"
href="/training"
subtitle="モデルのポストトレーニング"
>
現在パブリックプレビュー中の W&B Training を使用して、サーバーレスの強化学習( RL )を用いた大規模言語モデルのポストトレーニングを行います。フルマネージドの GPU インフラストラクチャー、 ART や RULER とのインテグレーション、マルチターンのエージェントタスク向けの自動スケーリングなどの機能が含まれます。
<br/>
<br/>
• <a href="/training">イントロダクション</a><br/>
• <a href="/training/prerequisites">クイックスタート</a><br/>
</ProductCard>
</div>
</HomeWrapper>
56 changes: 56 additions & 0 deletions ja/inference.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,56 @@
---
title: W&B Inference
description: W&B Weave と OpenAI 互換 API を通じて、オープンソースの基盤 モデル に アクセス する
mode: wide
---

W&B Inference を使用すると、W&B Weave および OpenAI 互換の API を通じて、主要なオープンソースの基盤 Models にアクセスできます。以下のことが可能です。

- ホスティングプロバイダーへの登録や自前での Model ホスティングを行うことなく、AI アプリケーションやエージェントを構築できます
- [W&B Weave Playground](/weave/guides/tools/playground) で [サポートされている Models](/inference/models) を試すことができます

Weave を使用すると、W&B Inference を活用したアプリケーションのトレース、評価、モニタリング、改善を行うことができます。


## クイックスタート

Python を使用した簡単な例を以下に示します。

```python
import openai

client = openai.OpenAI(
# カスタム base URL は W&B Inference を指します
base_url='https://api.inference.wandb.ai/v1',

# https://wandb.ai/settings で APIキー を作成してください
api_key="<your-api-key>",

# オプション: 使用状況を追跡するための Team と Project
project="<your-team>/<your-project>",
)

response = client.chat.completions.create(
model="meta-llama/Llama-3.1-8B-Instruct",
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Tell me a joke."}
],
)

print(response.choices[0].message.content)
```

## 次のステップ

1. [利用可能な Models](/inference/models) と [使用方法および制限](/inference/usage-limits/) を確認する
2. [事前準備](/inference/prerequisites/) ガイドに従ってアカウントを設定する
3. [API](/inference/api-reference/) または [UI](/inference/ui-guide/) を通じてサービスを利用する
4. [使用例](/inference/examples/) を試す

## 使用詳細

<Info>

料金、使用制限、クレジットに関する情報は、[使用方法および制限](/inference/usage-limits/) を参照してください。
</Info>
49 changes: 49 additions & 0 deletions ja/inference/api-reference.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,49 @@
---
title: API の概要
description: W&B Inference サービスに関する完全な API リファレンス
---

W&B Inference API を使用して、基盤モデルにプログラムからアクセスする方法を学びます。

## ベース URL

Inference サービスへのアクセス先:

```plaintext
https://api.inference.wandb.ai/v1
```

<Note>
**重要**

このエンドポイントを使用するには、以下が必要です。
- Inference クレジットを保有する W&B アカウント
- 有効な W&B APIキー

複数の Team に所属している場合や、使用量を特定の Project に紐付けたい場合は、Team ID と Project ID も必要になります。コードサンプルでは、これらは `<your-team>/<your-project>` として表示されます。指定しない場合は、デフォルトの Entity と `inference` というプロジェクト名が使用されます。
</Note>

## 利用可能なメソッド

W&B Inference API は、基盤モデルと対話するための OpenAI 互換のエンドポイントを提供します。

- **[Chat Completions](/inference/api-reference/chat-completions)** - さまざまな基盤モデルを使用してチャットの補完(completion)を作成します。
- **[List Models](/inference/api-reference/list-models)** - 利用可能なすべてのモデルとその ID を取得します。

## 認証

すべての API リクエストには、W&B APIキーによる認証が必要です。APIキーは [wandb.ai/settings](https://wandb.ai/settings) で作成してください。

リクエストヘッダーに APIキーを含めます:
- OpenAI SDK の場合: `api_key` パラメータとして設定します。
- 直接 API を呼び出す場合: `Authorization: Bearer <your-api-key>` を使用します。

## エラーハンドリング

エラーコードの完全なリストと解決方法については、 [API Errors](/inference/api-reference/errors) を参照してください。

## 次のステップ

- [使用例](/inference/examples) を試して、API の実際の動作を確認する。
- [UI](/inference/ui-guide) で Models を探索する。
- アカウントの [使用制限](/inference/usage-limits) を確認する。
90 changes: 90 additions & 0 deletions ja/inference/api-reference/chat-completions.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,90 @@
---
title: チャットコンプリーション
description: OpenAI 互換エンドポイントを使用して chat completions を作成する
---

`/chat/completions` エンドポイントを使用してチャットの補完(chat completion)を作成します。このエンドポイントは、メッセージの送信とレスポンスの受信に OpenAI フォーマットを採用しています。

## 要件

チャットの補完を作成するには、以下を提供してください:
- Inference サービスのベース URL: `https://api.inference.wandb.ai/v1`
- W&B APIキー: `<your-api-key>`
- オプション: W&B の Teams と Projects: `<your-team>/<your-project>`
- [利用可能な Models](/inference/models) リストにあるモデル ID

## リクエストの例

<Tabs>
<Tab title="Python">
```python
import openai

client = openai.OpenAI(
# カスタムベース URL は W&B Inference を指します
base_url='https://api.inference.wandb.ai/v1',

# https://wandb.ai/settings で APIキーを作成してください
# 安全のため、環境変数 OPENAI_API_KEY に設定することを検討してください
api_key="<your-api-key>",

# オプション: 使用状況を追跡するための Team と Project
project="<your-team>/<your-project>",
)

# <model-id> を利用可能なモデルリストの任意のモデル ID に置き換えてください
response = client.chat.completions.create(
model="<model-id>",
messages=[
{"role": "system", "content": "<your-system-prompt>"},
{"role": "user", "content": "<your-prompt>"}
],
)

print(response.choices[0].message.content)
```
</Tab>
<Tab title="Bash">
```bash
curl https://api.inference.wandb.ai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <your-api-key>" \
-H "OpenAI-Project: <your-team>/<your-project>" \
-d '{
"model": "<model-id>",
"messages": [
{ "role": "system", "content": "You are a helpful assistant." },
{ "role": "user", "content": "Tell me a joke." }
]
}'
```
</Tab>
</Tabs>

## レスポンス形式

API は OpenAI 互換の形式でレスポンスを返します:

```json
{
"id": "chatcmpl-...",
"object": "chat.completion",
"created": 1234567890,
"model": "meta-llama/Llama-3.1-8B-Instruct",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": "Here's a joke for you..."
},
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 25,
"completion_tokens": 50,
"total_tokens": 75
}
}
```
16 changes: 16 additions & 0 deletions ja/inference/api-reference/errors.mdx
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
---
title: API エラー
description: 一般的な API エラーとその解決方法
---

以下の表は、遭遇する可能性のある一般的なAPIエラーの一覧です。

| エラーコード | メッセージ | 原因 | 解決策 |
| ---------- | ------- | ----- | -------- |
| 401 | Authentication failed | 認証資格情報が正しくないか、W&B の Project 、 Entity 、または名称が間違っています。 | 正しい APIキー を使用していること、および W&B の Project 名と Entity が正しいことを確認してください。 |
| 402 | You exceeded your current quota, please check your plan and billing details | クレジットが不足しているか、月間の支出上限に達しました。 | クレジットを追加するか、制限を引き上げてください。詳細については、 [Usage information and limits](/inference/usage-limits/) を参照してください。 |
| 403 | Country, region, or territory not supported | サポートされていない場所からAPIにアクセスしています。 | [Geographic restrictions](/inference/usage-limits/#geographic-restrictions) を参照してください。 |
| 403 | The inference gateway is not enabled for your organization | 組織でインファレンスゲートウェイが有効化されていません。W&B Inference を使用するにはこれが必要です。 | 組織の W&B 管理者に連絡してインファレンスゲートウェイを有効にするか、 [W&B support](mailto:[email protected]) までお問い合わせください。 |
| 429 | Concurrency limit reached for requests | 同時リクエスト数が多すぎます。 | 同時リクエスト数を減らすか、制限を引き上げてください。詳細については、 [Usage information and limits](/inference/usage-limits/) を参照してください。 |
| 500 | The server had an error while processing your request | 内部 サーバー エラー。 | 少し待ってから再試行してください。問題が解決しない場合はサポートに連絡してください。 |
| 503 | The engine is currently overloaded, please try again later | サーバー のトラフィックが高負荷になっています。 | しばらく待ってからリクエストを再試行してください。 |
Loading
Loading