Clarification on JSON_OUTPUT field in text-generation-launcher command.

[ansSanthoshM]

Hi All,

I am setting the JSON_OUTPUT to True in text-generation-launcher command.
https://huggingface.co/docs/text-generation-inference/en/basic_tutorials/launcher#jsonoutput

**Should value of the --json-output be Boolean ? like below **

text-generation-launcher --model-id /data/$model --max-total-tokens 4096 --json-output true
In this case, what is the location of the json output file ?

OR
**Should value of the --json-output be file path ? like below **

text-generation-launcher --model-id /data/$model --max-total-tokens 4096 --json-output /home/$mode_log.json

[alexgravx]

According to the cli.py code in main, it should be Boolean. Here is the code snippet:

@app.command()
def serve(
    model_id: str,
    revision: Optional[str] = None,
    sharded: bool = False,
    quantize: Optional[Quantization] = None,
    speculate: Optional[int] = None,
    dtype: Optional[Dtype] = None,
    trust_remote_code: bool = False,
    uds_path: Path = "/tmp/text-generation-server",
    logger_level: str = "INFO",
    json_output: bool = False,
    otlp_endpoint: Optional[str] = None,
):
...