Instructions to use tokimoa/jp-invoice-reader-2b-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use tokimoa/jp-invoice-reader-2b-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Use Docker

docker model run hf.co/tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use tokimoa/jp-invoice-reader-2b-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tokimoa/jp-invoice-reader-2b-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tokimoa/jp-invoice-reader-2b-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Ollama
How to use tokimoa/jp-invoice-reader-2b-GGUF with Ollama:
```
ollama run hf.co/tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
```

Unsloth Studio

How to use tokimoa/jp-invoice-reader-2b-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for tokimoa/jp-invoice-reader-2b-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for tokimoa/jp-invoice-reader-2b-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for tokimoa/jp-invoice-reader-2b-GGUF to start chatting

How to use tokimoa/jp-invoice-reader-2b-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use tokimoa/jp-invoice-reader-2b-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use tokimoa/jp-invoice-reader-2b-GGUF with Docker Model Runner:
```
docker model run hf.co/tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M
```

Lemonade

How to use tokimoa/jp-invoice-reader-2b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.jp-invoice-reader-2b-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use tokimoa/jp-invoice-reader-2b-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default tokimoa/jp-invoice-reader-2b-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

tokimoa commited on 3 days ago

Commit

1502253

verified ·

1 Parent(s): 67eaa56

Upload folder using huggingface_hub

Browse files

Files changed (6) hide show

.gitattributes +4 -0
README.md +66 -0
jp-invoice-reader-2b-Q4_K_M.gguf +3 -0
jp-invoice-reader-2b-Q8_0.gguf +3 -0
jp-invoice-reader-2b-f16.gguf +3 -0
mmproj-jp-invoice-reader-2b-f16.gguf +3 -0

.gitattributes CHANGED Viewed

@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+jp-invoice-reader-2b-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+jp-invoice-reader-2b-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
+jp-invoice-reader-2b-f16.gguf filter=lfs diff=lfs merge=lfs -text
+mmproj-jp-invoice-reader-2b-f16.gguf filter=lfs diff=lfs merge=lfs -text

README.md ADDED Viewed

	@@ -0,0 +1,66 @@

+---
+license: apache-2.0
+language:
+- ja
+base_model: Qwen/Qwen3-VL-2B-Instruct
+pipeline_tag: image-text-to-text
+tags:
+- gguf
+- llama.cpp
+- japanese
+- document-understanding
+- invoice
+---
+# jp-invoice-reader-2b-GGUF — 日本語帳票リーダー
+日本語帳票（請求書・見積書・納品書・領収書）の読み取りに特化した [jp-invoice-reader-2b](https://huggingface.co/tokimoa/jp-invoice-reader-2b) のGGUF版です。llama.cppでWindows/LinuxのCPU・GPUで動きます。
+## ファイル
+| ファイル | サイズ | 用途 |
+|---|---|---|
+| **jp-invoice-reader-2b-Q8_0.gguf** | 1.7GB | 推奨（抽出・QAとも） |
+| jp-invoice-reader-2b-Q4_K_M.gguf | 1.0GB | QA用途向け（下記注意） |
+| jp-invoice-reader-2b-f16.gguf | 3.2GB | 参照用 |
+| mmproj-jp-invoice-reader-2b-f16.gguf | 781MB | 視覚エンコーダ（**必須・どの量子化とも組み合わせ可**） |
+## 実測（公開中のbf16版と同一プロトコル: 未見評価100枚・QA400問・抽出700フィールド）
+| 指標 | bf16（MLX公開版） | Q8_0 | Q4_K_M |
+|---|---|---|---|
+| ja-docs QA（実写風6問） | 6/6 | **6/6** | 6/6 |
+| 帳票QA | 93.5% | **91.8%** | 90.5% |
+| 抽出フィールド一致（厳密） | 90.1% | **84.3%** | 67.3% |
+厳密一致の低下について実態を補足します（同一100枚で追加検証）:
+- **Q8_0の低下分はほぼ形式ゆれです。** 消費税額を`{"rate": 10, "amount": 1449000}`のようにネストで返す揺れが増え、厳密一致では×になりますが、**税額の値ベースでは99/100が正解**でした。金額の読み取り能力自体は維持されています。後段でJSONを受けるときはネスト形式も許容してください。
+- **Q4_K_MはJSON構文の崩れが100枚中15件発生します。** 抽出（長い構造化出力）には非推奨で、QA用途向けです。
+## 使い方
+```bash
+llama-server -m jp-invoice-reader-2b-Q8_0.gguf --mmproj mmproj-jp-invoice-reader-2b-f16.gguf \
+  --port 8080 --jinja -c 8192 --image-min-tokens 64 --image-max-tokens 4096
+```
+運用上の注意（いずれも実測に基づく）:
+1. **APIでは画像をテキストより先に置いてください**（contentの並びを`[image_url, text]`にする）。逆順は精度が下がります。
+2. **`-c 8192`等でコンテキスト長を指定してください。** 無指定はモデル宣言の262Kで確保しようとしてメモリ不足になります。
+3. **`--image-max-tokens 4096`を推奨します。** 帳票は高解像度のため、既定では画像が過剰に縮小され数値の読み取りが落ちます。
+プロンプト例: 抽出は「この帳票から主要項目を抽出してJSONで出力してください。」、QAは「この請求書の合計金額は？」等の自由質問です。
+## 制約
+合成帳票ドメインで学習しています。実写スキャン・手書き・特殊レイアウトは未検証で、写真由来の帳票では精度が下がる可能性があります。
+## ライセンスと帰属
+- ベース: [Qwen/Qwen3-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-2B-Instruct)（Apache-2.0）
+- 学習データ・評価方法の詳細は[本体repo](https://huggingface.co/tokimoa/jp-invoice-reader-2b)を参照してください
+---
+Developed by [tokimoa](https://tokimoa.jp) — データを外に出さないAI活用を支援しています

jp-invoice-reader-2b-Q4_K_M.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:4706a2c4c942f2ecbc761dbb9ca0d8378607fcfed7f4f4f1b69bad6657cc8f6b
+size 1107410336

jp-invoice-reader-2b-Q8_0.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:586dada4c6ab3b6a6cb6e35d54c7d065afcd0829d905f7edb936a500fc02ec14
+size 1834427808

jp-invoice-reader-2b-f16.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7754cf026bbb98cdfae403bf97ec4c518b336d22c75e6e4afb9b5e31e18409ee
+size 3447350688

mmproj-jp-invoice-reader-2b-f16.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:43857c950a6fc1726f573d3dde87b3d08835c456493139b634a3e8227078b6c3
+size 819394752