Skip to content

Index

`forking` ¶

`AnalyzeScratchbubbles(spec: ExecutionSpec, base: Optional[Node] = None)` ¶

Bases: ForkAnalysis

Analyze fork behavior in a Scratchbubbles checkpoint.

`AnalyzeThoughtbubbles(spec: ExecutionSpec, base: Optional[Node] = None)` ¶

Bases: ForkAnalysis

Analyze fork behavior in a regular Thoughtbubbles checkpoint.

`AnalyzeScratchSelector(spec: ExecutionSpec, base: Optional[Node] = None)` ¶

Bases: AnalyzeScratchbubbles

Capture selector probabilities, independently of the transformer attention.

`measure_selector(weights: np.ndarray, query_positions: np.ndarray, source_positions: np.ndarray, source_scores: np.ndarray, token_ids: list[int], token_text: list[str], top_k: int) -> dict[str, Any]` `staticmethod` ¶

Keep key-stream probabilities and pool their mass by original token.

`FullPrefixEvaluator(spec: ExecutionSpec, base: Node | None = None)` ¶

Bases: Evaluator[M], Generic[M]

Evaluate cacheless forking models by recomputing the visible prefix.

`DiversityScratchbubbles` ¶

Bases: Scratchbubbles

Scratchbubbles training model with fork-diversity instrumentation.

`measure_friend_diversity(x: jax.Array, token_index: jax.Array) -> jax.Array` `staticmethod` ¶

Measure mean cosine distance between non-rightmost siblings.

`decode(x: jax.Array, padding_mask: Optional[jax.Array] = None, deterministic: bool = False, **kwargs: Any) -> Any` ¶

Decode while recording states immediately after each fork.