* Create a new tensor (shape in reverse order). * @param {string} type - Tensor type. * @param {...number} dims - Tensor shape.

* Set value at the given index. * @param {number} index * @param {number} value * @returns {Tensor}

* Load a model from a file. * @type {<T extends typeof Model>(this: T, path: string, hparams?: ConstructorParameters<T>[1]) => InstanceType<T>}

* Get value at the given index. * @param {number} index * @returns {number}

* Decode one token into a Uint8Array. * @param {number} token * @returns {Uint8Array} * @abstract

* Encode a string into tokens. * @param {string} text * @returns {number[]} * @abstract

* Compute the graph.

* @param {number[]} input - Array of input token IDs. * @param {object} [options] * @param {number} [options.max_tokens=100] - Maximum number of

* Initialize the context.

* Build a graph for forward computation. * @param {Tensor} tensor

* Go through all tensors in the model and collect them with their names. * @returns {Array<[string, Tensor]>}

* Split a tensor into chunks of the given size along the first dimension. * @param {number} chunkSize * @returns {Tensor[]}

* Get a view of the tensor (shape in reverse order). * @param {...number} dims * @returns {Tensor}

* TODO: document this

* Make the tensor contiguous in memory. * @returns {Tensor}

* Decode tokens into a string. * @param {number[]} tokens * @returns {string}

* @override

* @param {Tensor} x - Input tensor. * @param {any} state * @param {Tensor[]} updates - Array of tensors that should be updated after the forward

* Prepare the initial state. * @param {Context} cx * @returns {any} Initial state. * @abstract

* Permute the tensor. * @param {number} axis0 * @param {number} axis1 * @param {number} axis2 * @param {number} axis3 * @returns {Tenso

* Print all tensors in the model.

* Set all elements to the given value. * @param {number} value * @returns {Tensor}

* Create an empty model. * @param {Context} context * @param {Object} hparams

* @param {Module} parentModule - A parent module. * @param {number} numEmbeddings - The size of the vocabulary. * @param {number} dim - The embe

* @param {Module} parentModule - A parent module. * @param {number} dim - The embedding dimension.

* @param {Module} parentModule - A parent module. * @param {number} inputDim - The input dimension. * @param {number} outputDim - The output dim

* Create a new module. * @param {Module} parentModule

* Create a new RWKV model. * @param {Context} context * @param {{ vocab_size: number, embed_dim: number, num_layers: number }} hparams

* @override

* Size of one element in bytes. * @returns {BigInt}

* Load a tokenizer from a file. * @type {<T extends typeof Tokenizer>(this: T, path: string) => InstanceType<T>}

* Print info about the graph.

* Print GGML objects

* Reset the graph.

* Reshape the tensor. * @param {...number} dims * @returns {Tensor}

* Tensor shape (in reverse order). * @returns {number[]}

* Debug string.

* Transpose the tensor. * @returns {Tensor}

* Tensor type. * @returns {string}