Functions — github.com/OpenPipe/ART

hub / github.com/OpenPipe/ART / functions

Functions1,502 in github.com/OpenPipe/ART

⨍ Functions 1,502◇ Types & classes 261↳ Endpoints 14

↓ 1 callersMethod_run

(self, host: str, port: int)

src/art/tinker/server.py:123

↓ 1 callersMethod_run_openai_server

( self, state: ModelState, host: str, port: int, )

src/art/tinker_native/backend.py:372

↓ 1 callersMethod_save_checkpoint

( self, training_client: tinker.TrainingClient, checkpoint_name: str, )

src/art/tinker_native/backend.py:783

↓ 1 callersFunction_schema_comments

Extract comments from schema (title, description, examples).

src/art/tinker/cookbook_v/renderers/gpt_oss.py:84

↓ 1 callersMethod_should_eval_step

(self, step: int)

src/art/pipeline_trainer/trainer.py:761

↓ 1 callersMethod_skip_scenarios

( self, scenarios: AsyncIterator[ScenarioT], count: int )

src/art/pipeline_trainer/trainer.py:280

↓ 1 callersFunction_slugify

Create a filesystem-friendly slug from text.

dev/yes-no-maybe-vision/generate_images.py:229

↓ 1 callersMethod_start_vllm_subprocess

Launch vLLM as a subprocess on inference GPUs. Returns (host, port).

src/art/unsloth/service.py:333

↓ 1 callersMethod_status_loop

(self)

src/art/pipeline_trainer/trainer.py:573

↓ 1 callersFunction_strip_logprobs

(obj: Any)

src/art/utils/strip_logprobs.py:35

↓ 1 callersMethod_tinker_sample_call

(self, label: str, awaitable: Awaitable[T])

src/art/tinker_native/backend.py:170

↓ 1 callersFunction_to_async_iterator

Convert a sync Iterable to an AsyncIterator, or pass through if already async.

src/art/pipeline_trainer/trainer.py:23

↓ 1 callersMethod_to_openai_tool_arguments

Convert tool arguments for OpenAI-compatible message payloads. Qwen3 templates accept JSON-string arguments directly; subclasses can

src/art/tinker/cookbook_v/renderers/qwen3.py:286

↓ 1 callersFunction_tool_name_and_args

(tool_call: Any)

src/art/pipeline_trainer/binary_prefix_tool_pipeline.py:127

↓ 1 callersMethod_train_dedicated

Train in dedicated mode — no sleep/wake, vLLM keeps running on separate GPU.

src/art/unsloth/service.py:579

↓ 1 callersMethod_train_model

( self, model: AnyTrainableModel, trajectory_groups: list[TrajectoryGroup], co

src/art/serverless/backend.py:321

↓ 1 callersMethod_train_model

( self, model: TrainableModel, trajectory_groups: list[TrajectoryGroup], confi

src/art/local/backend.py:687

↓ 1 callersMethod_train_sft

( self, model: AnyTrainableModel, trajectories: Iterable[Trajectory], config:

src/art/backend.py:54

↓ 1 callersMethod_train_shared

Train in shared mode — sleep/wake cycle with in-process vLLM.

src/art/unsloth/service.py:640

↓ 1 callersFunction_trajectory_has_choice

(trajectory: Trajectory)

src/art/tinker_native/data.py:120

↓ 1 callersMethod_trigger_collapse

(self)

src/art/pipeline_trainer/trainer.py:697

↓ 1 callersFunction_unflatten_message

Convert flat parquet format back to message dict.

src/art/utils/trajectory_logging.py:46

↓ 1 callersMethod_update_throughput_metrics

(self, result: dict[str, float])

src/art/metrics.py:376

↓ 1 callersFunction_upload_model

Uploads a model to Together.

src/art/utils/deployment/together.py:82

↓ 1 callersFunction_use_adapter

Context manager that switches to a named LoRA adapter, then restores the original.

src/art/unsloth/train.py:301

↓ 1 callersMethod_validate_and_add

(self, key: str, value: float)

src/art/metrics.py:319

↓ 1 callersFunction_wait_for_count

(values: list[int], target: int, timeout: float = 2.0)

tests/unit/test_mp_actors.py:53

↓ 1 callersFunction_wait_for_job

Waits for a model deployment job to complete in Together.

src/art/utils/deployment/together.py:173

↓ 1 callersMethod_wait_for_policy

(self)

src/art/pipeline_trainer/trainer.py:309

↓ 1 callersMethod_wait_for_server_ready

( self, base_url: str, api_key: str, model: TrainableModel )

src/art/tinker_native/backend.py:506

↓ 1 callersMethod_wrap_qwen_tool_response

Wrap tool response content in Qwen's <tool_response> tags.

src/art/tinker/cookbook_v/renderers/qwen3.py:126

↓ 1 callersMethod_wrap_qwen_tool_response_chunks

Wrap content chunks in Qwen's <tool_response> tags for multimodal messages.

src/art/tinker/cookbook_v/renderers/qwen3.py:489

↓ 1 callersMethod_write_log_line

(self, line: str)

src/art/pipeline_trainer/status.py:311

↓ 1 callersFunctionadd_thread

(thread_id, base_url, api_key, model)

src/art/langgraph/llm_wrapper.py:26

↓ 1 callersMethodaincrement

(self)

tests/unit/test_mp_actors.py:20

↓ 1 callersMethodainvoke

(self, input, config=None, **kwargs)

src/art/langgraph/llm_wrapper.py:164

↓ 1 callersFunctionapply_agent_move

( game: TicTacToeGame, square: str, symbol: Literal["x", "o"] )

examples/tic_tac_toe_self_play/game_utils.py:53

↓ 1 callersFunctionapply_agent_move

(game: TicTacToeGame, move: str)

examples/tic_tac_toe/game_utils.py:47

↓ 1 callersFunctionapply_lora_adapters

( model: Sequence[torch.nn.Module], provider: GPTModelProvider, )

src/art/megatron/lora.py:369

↓ 1 callersFunctionarchive_and_presign_step_url

Get a presigned URL for a step in a model. Args: model_name: Name of the model. project: Project name. step: Step number.

src/art/utils/s3.py:266

↓ 1 callersFunctionarun

()

examples/mcp-rl/servers/python/mcp_balldontlie/server.py:307

↓ 1 callersFunctionarun

()

examples/mcp-rl/servers/python/mcp_googlemaps/server.py:475

↓ 1 callersFunctionarun

()

examples/mcp-rl/servers/python/mcp_alphavantage/server.py:348

↓ 1 callersMethodbind_tools

(self, tools)

src/art/langgraph/llm_wrapper.py:199

↓ 1 callersFunctionbuild_cost_calculator

Return a callable that computes split-scoped Tinker inference costs.

src/art/costs.py:113

↓ 1 callersFunctionbuild_datum

( prompt_tokens: list[int], completion_tokens: list[int], logprobs: list[float], advantage: fl

src/art/tinker_native/data.py:241

↓ 1 callersMethodbuild_generation_prompt

Build generation prompt with <think> prefill for thinking mode.

src/art/tinker/cookbook_v/renderers/kimi_k25.py:94

↓ 1 callersFunctionbuild_internal_config

()

dev/yes-no-maybe-metrics.py:158

↓ 1 callersFunctionbuild_prompts

()

dev/yes-no-maybe-metrics.py:48

↓ 1 callersFunctionbuild_s3_zipped_step_path

Return the fully-qualified S3 URI for a zipped step in a model directory.

src/art/utils/s3.py:47

↓ 1 callersFunctionbuild_scenarios

Generate all scenario variations.

src/art/pipeline_trainer/yes_no_maybe_pipeline.py:30

↓ 1 callersMethodbuild_supervised_example

Build tokens and per-token weights for supervised fine-tuning. This default implementation concatenates rendered messages in order.

src/art/tinker/cookbook_v/renderers/base.py:1028

↓ 1 callersFunctionbump_version

Bump version based on type (major, minor, patch).

scripts/bump_version.py:27

↓ 1 callersFunctioncalculate_attn_bias

( batch_size: int, seq_len: int, device: torch.device, group_ids: torch.Tensor, parent_ids

src/art/unsloth/train.py:248

↓ 1 callersFunctioncalculate_mask

( batch_size: int, seq_len: int, device: torch.device, group_ids: torch.Tensor, parent_ids

src/art/unsloth/train.py:275

↓ 1 callersFunctioncall_score_title

Async wrapper for scoring.

examples/hn_title_generator/train.py:115

↓ 1 callersMethodchat_completion_and_token_discrepancies

( self, base_model: str, sample_response: tinker.SampleResponse, model_name: s

src/art/tinker/server.py:402

↓ 1 callersFunctioncheck_game_finished

(game: TwentyFortyEightGame)

examples/2048/utils.py:167

↓ 1 callersFunctioncheck_hallucinated_facts

Check if the summary hallucinates facts that are not in the original text

examples/just-the-facts/just_the_facts/checks.py:65

↓ 1 callersFunctioncheck_has_conservative_bias

Check if the summary hallucinates facts that are not in the original text

examples/just-the-facts/just_the_facts/checks.py:110

↓ 1 callersFunctioncheck_has_liberal_bias

Check if the summary has a liberal bias

examples/just-the-facts/just_the_facts/checks.py:151

↓ 1 callersFunctioncheck_if_titles_match_bodies

(bodies, titles)

examples/hn_title_generator/reference_grpo_trainer.py:148

↓ 1 callersFunctioncheck_includes_all_facts

Check if the summary includes all of the facts from the original text

examples/just-the-facts/just_the_facts/checks.py:20

↓ 1 callersFunctioncheck_successful

Check if the agent completed the task successfully using GPT-4 evaluation. Args: trajectory: The trajectory containing messages, tool cal

examples/mcp-rl/mcp_rl/checks.py:24

↓ 1 callersFunctioncheck_title_matches_body

( client: openai.AsyncOpenAI, body: str, title: str )

examples/hn_title_generator/train.py:120

↓ 1 callersMethodchild_exit

(self)

tests/unit/test_mp_actors.py:44

↓ 1 callersMethodclose

(self)

src/art/pipeline_trainer/status.py:80

↓ 1 callersFunctioncompute_advantages

( rewards: list[float], normalize_advantages: bool = True )

src/art/tinker_native/data.py:23

↓ 1 callersFunctioncompute_train_cost

Compute training cost from token count.

src/art/costs.py:131

↓ 1 callersFunctioncondense_board

(game: TwentyFortyEightGame, direction: Direction)

examples/2048/utils.py:111

↓ 1 callersFunctionconsume

()

tests/unit/test_mp_actors.py:96

↓ 1 callersFunctionconvert_fused_moe_lora

Convert fused MoE LoRA tensors to per-expert format. Non-expert tensors (e.g. self_attn) are passed through unchanged.

src/art/utils/convert_moe_lora.py:69

↓ 1 callersFunctionconvert_langgraph_messages

(messages: List[object])

src/art/langgraph/message_utils.py:90

↓ 1 callersFunctioncreate_chat_completion

( client: openai.AsyncOpenAI, *, model_name: str, messages: art.Messages, max_tokens: int,

dev/yes-no-maybe-metrics.py:28

↓ 1 callersFunctioncreate_conversation_prefix_with_tools

Create conversation prefix with tools using the renderer implementation.

src/art/tinker_native/data.py:16

↓ 1 callersMethodcreate_conversation_prefix_with_tools

Create message(s) with tool specifications to prepend to conversations. Returns one or more messages to prepend to the conversation. This is

src/art/tinker/cookbook_v/renderers/base.py:896

↓ 1 callersFunctioncreate_messages_from_logs

(log_path: str, trajectory: Trajectory)

src/art/langgraph/llm_wrapper.py:40

↓ 1 callersFunctioncreate_sample_resources

Create sample resources for testing.

examples/mcp-rl/test_scenario_generation.py:85

↓ 1 callersFunctioncreate_shared_prefix_attention_state

Build a compiled block mask for ART shared-prefix packing. Initialized on the device of the group_ids tensor. Args: group_ids: `[B,

src/art/megatron/flex_attention.py:69

↓ 1 callersFunctiondelete_checkpoint

( checkpoint_dir: Path, rest_client: TinkerRestClient )

src/art/tinker/service.py:283

↓ 1 callersFunctiondelete_checkpoints

(output_dir: str, excluding: list[int])

src/art/local/checkpoints.py:7

↓ 1 callersMethoddelete_checkpoints

(self, steps_to_keep: list[int])

src/art/tinker/service.py:172

↓ 1 callersFunctiondemo_basic_logging

Demonstrate the basic logging functions.

dev/demo_logging.py:9

↓ 1 callersFunctiondemo_color_codes

Demonstrate the color code constants.

dev/demo_logging.py:25

↓ 1 callersFunctiondemo_error_scenarios

Demonstrate error reporting scenarios.

dev/demo_logging.py:123

↓ 1 callersFunctiondemo_formatting_combinations

Demonstrate various formatting combinations.

dev/demo_logging.py:146

↓ 1 callersFunctiondemo_log_levels

Demonstrate different log levels in action.

dev/demo_logging.py:177

↓ 1 callersFunctiondemo_progress_tracking

Demonstrate progress tracking with logging.

dev/demo_logging.py:99

↓ 1 callersFunctiondemo_real_world_usage

Demonstrate real-world usage scenarios.

dev/demo_logging.py:65

↓ 1 callersFunctiondemo_timestamp

Demonstrate the timestamp function.

dev/demo_logging.py:52

↓ 1 callersFunctiondeploy

()

scripts/deploy-model.py:50

↓ 1 callersFunctiondeploy_step

()

examples/tic_tac_toe_self_play/deploy_step.py:12

↓ 1 callersFunctiondeploy_to_together

Deploy a model checkpoint to Together. Args: model: The TrainableModel to deploy. checkpoint_path: Local path to the checkpoint d

src/art/utils/deployment/together.py:189

↓ 1 callersFunctiondeploy_wandb

Deploy a model to W&B by uploading a LoRA artifact. Args: model: The TrainableModel to deploy. checkpoint_path: Local path to the

src/art/utils/deployment/wandb.py:36

↓ 1 callersFunctiondeserialize_trajectory_groups

Deserialize trajectory groups from a JSONL or YAML string.

src/art/utils/trajectory_migration.py:97

↓ 1 callersFunctiondetect_chat_template_parts

Detect instruction and response parts from a tokenizer's chat template. Args: tokenizer: A tokenizer with a chat_template attribute

src/art/utils/model_config.py:25

↓ 1 callersFunctiondict_to_message_or_choice

(d: dict[str, Any])

src/art/utils/trajectory_migration.py:132

↓ 1 callersFunctiondict_to_trajectory

(d: dict[str, Any])

src/art/utils/trajectory_migration.py:119

↓ 1 callersFunctiondict_to_trajectory_group

(d: dict[str, Any])

src/art/utils/trajectory_migration.py:110

↓ 1 callersFunctionencode_tools_to_typescript_style

Convert tools (list of dict) to TypeScript style string. Supports OpenAI format: {"type": "function", "function": {...}} Args:

src/art/tinker/cookbook_v/renderers/kimi_k2_5_tool_declaration_ts.py:461

← previous next →501–600 of 1,502, ranked by callers