maxent_grpo.training.telemetry.wandb¶

Minimal Weights & Biases environment setup helpers.

The helpers here expose the W&B entity/project/group stored in training arguments as WANDB_* environment variables so any downstream code that initializes W&B picks them up automatically.

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the specific language governing permissions and limitations under the License.

Functions

init_wandb_training(training_args)

Initialize Weights & Biases environment variables for a run.

Classes

WandbConfig(*args, **kwargs)

Protocol for objects with W&B configuration attributes.

class maxent_grpo.training.telemetry.wandb.WandbConfig(*args, **kwargs)[source]¶

Bases: Protocol

Protocol for objects with W&B configuration attributes.

wandb_entity: str | None¶

wandb_project: str | None¶

wandb_run_group: str | None¶

maxent_grpo.training.telemetry.wandb.init_wandb_training(training_args)[source]¶

Initialize Weights & Biases environment variables for a run.

Exposes entity/project/group from training_args to the W&B backend via WANDB_* environment variables. Values are only set when present on the provided config; missing attributes leave existing environment variables untouched.

Parameters:: training_args (WandbConfig) – Training configuration providing wandb_entity, wandb_project, and wandb_run_group fields.
Returns:: None. Environment variables are set as a side effect.
Return type:: None