timesfm-ts - v1.0.0
    Preparing search index...

    Interface PreprocessedData

    Output of the full preprocessing pipeline — ready for model inference.

    interface PreprocessedData {
        patchedInputs: Float32Array<ArrayBufferLike>[];
        patchedMasks: Uint8Array<ArrayBufferLike>[];
        contextMu: Float32Array<ArrayBufferLike>[];
        contextSigma: Float32Array<ArrayBufferLike>[];
        lastStats: RunningStats[];
        numPatches: number;
        batchSize: number;
        cleanedInputs: Float32Array<ArrayBufferLike>[];
        truncatedInputs: Float32Array<ArrayBufferLike>[];
    }
    Index

    Properties

    patchedInputs patchedMasks contextMu contextSigma lastStats numPatches batchSize cleanedInputs truncatedInputs

    Properties

    patchedInputs: Float32Array<ArrayBufferLike>[]

    Patched input [batchSize][numPatches * inputPatchLen] — flat per batch entry.

    patchedMasks: Uint8Array<ArrayBufferLike>[]

    Patch-level mask [batchSize][numPatches * inputPatchLen].

    contextMu: Float32Array<ArrayBufferLike>[]

    Per-patch means [batchSize * numPatches] — for RevIN reversal.

    contextSigma: Float32Array<ArrayBufferLike>[]

    Per-patch std deviations [batchSize * numPatches] — for RevIN reversal.

    lastStats: RunningStats[]

    Per-batch-element running stats after the last patch.

    numPatches: number

    Number of patches per series (= maxContext / inputPatchLen).

    batchSize: number

    Number of input series.

    cleanedInputs: Float32Array<ArrayBufferLike>[]

    The cleaned raw inputs (for post-processing reference).

    truncatedInputs: Float32Array<ArrayBufferLike>[]

    The truncated inputs (after truncation to maxContext, before padding). Reserved for future iterative refinement APIs.

    Settings

    Member Visibility

    On This Page

    Properties