Convert some of the models linear layers to lora layers. Args: model (nn.Module): The neural network model. num_layers (int)

Process prompt tokens through the model. Args: tokens: List of token sequences to process.

Save the model configuration to the ``config_path``. The final configuration will be sorted before saving for better readability. Args:

Inspired by: - stackoverflow.com/a/8356620/593036 - stackoverflow.com/questions/9535954/printing-lists-as-tabular-data

Convert parameter value based on its type in the schema.

Extract all possible types from a JSON schema definition. Handles anyOf, oneOf, allOf, type arrays, and enum fields. Args: schem

Calculates intermediate size based on multiplier, rounding up to multiple of 256.

Finds the smallest multiple of k greater than or equal to n.

Ops-based reference implementation for a single recurrent step. Shapes: - q, k: [B, H, Dk] - v: [B, H, Dv] - g: [B, H] or

Convert Gemma 4 tool call args to valid JSON. Gemma 4 uses unquoted keys and <|"|> as string delimiters instead of standard double quotes.

Extract argument configuration for a function.

Attempt to auto-infer a tool parser from the chat template.

Truncate the prefix of the string after the first occurrence of pattern.