hiyouga/LlamaFactory · warning · NotImplementedError

Environment information is not implemented yet.

Error message

Environment information is not implemented yet.

What it means

The v1 launcher dispatches CLI subcommands to v1 implementations, but the 'env' command has no v1 implementation yet, so it raises NotImplementedError. It is a known gap in the experimental v1 architecture, not a broken environment.

Source

Thrown at src/llamafactory/v1/launcher.py:134

            torchrun_args + script_args,
            env=env,
            check=True,
        )

        sys.exit(process.returncode)

    elif command == "chat":
        from .samplers.cli_sampler import run_chat

        run_chat()

    elif command == "merge":
        from llamafactory.v1.plugins.model_plugins.peft import merge_and_export_model

        merge_and_export_model()

    elif command == "env":
        raise NotImplementedError("Environment information is not implemented yet.")

    elif command == "version":
        raise NotImplementedError("Version information is not implemented yet.")

    elif command == "help":
        print(USAGE)

    elif command in _DIST_TRAIN_COMMANDS:
        # Single GPU training without torchrun
        if command in ("train", "sft"):
            from llamafactory.v1.trainers.sft_trainer import run_sft

            run_sft()
        elif command == "dpo":
            from llamafactory.v1.trainers.dpo_trainer import run_dpo

            run_dpo()
        elif command == "rm":

View on GitHub (pinned to f28afaf635)

Solutions

1. Unset USE_V1 and run `llamafactory-cli env` under v0.
2. Gather environment info manually: `pip list | grep -E 'torch|transformers|deepspeed'` and `nvidia-smi`.
3. Track/watch the v1 launcher for the implementation landing.

Example fix

# before
USE_V1=1 llamafactory-cli env

# after
unset USE_V1
llamafactory-cli env

import os
assert os.environ.get("USE_V1") != "1" or command != "env", "'env' is unimplemented in v1; unset USE_V1"

try:
    run_cli("env")
except NotImplementedError:
    os.environ.pop("USE_V1", None)
    run_cli("env")

When it happens

Trigger: Running `llamafactory-cli env` (or `lmf env`) with USE_V1=1 set in the environment.

Common situations: User switches to the experimental v1 path to try new trainers, then runs the familiar `env` diagnostics command from v0 docs.

Related errors

• Version information is not implemented yet.
• Batch infer is not implemented.
• Invalid role
• SGLang only supports n=1.
• SGLang engine does not support `get_scores`.

AI-assisted analysis of hiyouga/LlamaFactory@f28afaf635 (2026-08-14). Data as JSON: /api/errors/1e9e1ef73c88092a. Report an issue: GitHub.