Vllm.Tokenizers (VLLM v0.3.0)

Copy Markdown View Source

Submodule bindings for vllm.tokenizers.

Version

  • Requested: 0.14.0
  • Observed at generation: 0.14.0

Runtime Options

All functions accept a __runtime__ option for controlling execution behavior:

Vllm.Tokenizers.some_function(args, __runtime__: [timeout: 120_000])

Supported runtime options

  • :timeout - Call timeout in milliseconds (default: 120,000ms / 2 minutes)
  • :timeout_profile - Use a named profile (:default, :ml_inference, :batch_job, :streaming)
  • :stream_timeout - Timeout for streaming operations (default: 1,800,000ms / 30 minutes)
  • :session_id - Override the session ID for this call
  • :pool_name - Target a specific Snakepit pool (multi-pool setups)
  • :affinity - Override session affinity (:hint, :strict_queue, :strict_fail_fast)

Timeout Profiles

  • :default - 2 minute timeout for regular calls
  • :ml_inference - 10 minute timeout for ML/LLM workloads
  • :batch_job - Unlimited timeout for long-running jobs
  • :streaming - 2 minute timeout, 30 minute stream_timeout

Example with timeout override

# For a long-running ML inference call
Vllm.Tokenizers.predict(data, __runtime__: [timeout_profile: :ml_inference])

# Or explicit timeout
Vllm.Tokenizers.predict(data, __runtime__: [timeout: 600_000])

# Route to a pool and enforce strict affinity
Vllm.Tokenizers.predict(data, __runtime__: [pool_name: :strict_pool, affinity: :strict_queue])

See SnakeBridge.Defaults for global timeout configuration.

Summary

Functions

__all__()

Python module attribute vllm.tokenizers.__all__.

cached_get_tokenizer(tokenizer_name, opts \\ [])

Gets a tokenizer for the given model name via HuggingFace or ModelScope.

cached_tokenizer_from_config(model_config, opts \\ [])

Python binding for vllm.tokenizers.cached_tokenizer_from_config.

get_tokenizer(tokenizer_name, opts \\ [])

Gets a tokenizer for the given model name via HuggingFace or ModelScope.

tokenizer_registry()

Python module attribute vllm.tokenizers.TokenizerRegistry.

Functions

__all__()

@spec __all__() :: {:ok, [term()]} | {:error, Snakepit.Error.t()}

Python module attribute vllm.tokenizers.__all__.

Returns

  • list(term())

cached_get_tokenizer(tokenizer_name, opts \\ [])

@spec cached_get_tokenizer(
  term(),
  keyword()
) :: {:ok, term()} | {:error, Snakepit.Error.t()}

Gets a tokenizer for the given model name via HuggingFace or ModelScope.

Parameters

  • tokenizer_name (term())
  • args (term())
  • tokenizer_cls (term() keyword-only default: <class 'vllm.tokenizers.protocol.TokenizerLike'>)
  • trust_remote_code (boolean() keyword-only default: False)
  • revision (term() keyword-only default: None)
  • download_dir (term() keyword-only default: None)
  • kwargs (term())

Returns

  • term()

cached_tokenizer_from_config(model_config, opts \\ [])

@spec cached_tokenizer_from_config(
  term(),
  keyword()
) :: {:ok, nil} | {:error, Snakepit.Error.t()}

Python binding for vllm.tokenizers.cached_tokenizer_from_config.

Parameters

  • model_config (term())

Returns

  • nil

get_tokenizer(tokenizer_name, opts \\ [])

@spec get_tokenizer(
  term(),
  keyword()
) :: {:ok, term()} | {:error, Snakepit.Error.t()}

Gets a tokenizer for the given model name via HuggingFace or ModelScope.

Parameters

  • tokenizer_name (term())
  • args (term())
  • tokenizer_cls (term() keyword-only default: <class 'vllm.tokenizers.protocol.TokenizerLike'>)
  • trust_remote_code (boolean() keyword-only default: False)
  • revision (term() keyword-only default: None)
  • download_dir (term() keyword-only default: None)
  • kwargs (term())

Returns

  • term()

tokenizer_registry()

@spec tokenizer_registry() :: {:ok, term()} | {:error, Snakepit.Error.t()}

Python module attribute vllm.tokenizers.TokenizerRegistry.

Returns

  • term()