Instructions to use anthonym21/json-tokenizer-structured with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use anthonym21/json-tokenizer-structured with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="anthonym21/json-tokenizer-structured")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("anthonym21/json-tokenizer-structured", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use anthonym21/json-tokenizer-structured with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "anthonym21/json-tokenizer-structured"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anthonym21/json-tokenizer-structured",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/anthonym21/json-tokenizer-structured

SGLang

How to use anthonym21/json-tokenizer-structured with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "anthonym21/json-tokenizer-structured" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anthonym21/json-tokenizer-structured",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "anthonym21/json-tokenizer-structured" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "anthonym21/json-tokenizer-structured",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use anthonym21/json-tokenizer-structured with Docker Model Runner:
```
docker model run hf.co/anthonym21/json-tokenizer-structured
```

anthonym21 commited on Mar 5

Commit

d54e053

verified ·

1 Parent(s): 8ae05a9

Upload json_tokenizer/init.py with huggingface_hub

Browse files

Files changed (1) hide show

json_tokenizer/__init__.py +24 -0

json_tokenizer/__init__.py ADDED Viewed

	@@ -0,0 +1,24 @@

+"""
+json_tokenizer — A tokenizer optimized for JSON structures.
+Architecture:
+  - Structural tokens: single-token representations for JSON grammar ({, }, [, ], :, ,)
+  - Key tokens: deduplicated key vocabulary with Key() wrapper
+  - Value BPE: byte-pair encoding trained on JSON string/number values
+  - Type tokens: explicit type markers for faithful roundtrip encoding
+Delivers 5-15% fewer tokens than cl100k_base on schema-repetitive JSON
+with a 90x smaller vocabulary, and lossless roundtrip fidelity.
+"""
+from json_tokenizer.tokenizer import JSONTokenizer
+from json_tokenizer.bpe import BPETrainer
+__version__ = "0.2.0"
+__all__ = ["JSONTokenizer", "BPETrainer"]
+try:
+    from json_tokenizer.hf_compat import JSONPreTrainedTokenizer
+    __all__.append("JSONPreTrainedTokenizer")
+except ImportError:
+    pass  # transformers not installed

Upload json_tokenizer/__init__.py with huggingface_hub

Upload json_tokenizer/init.py with huggingface_hub