continuum::backend::ConformanceOptions Struct Reference

Inputs the kit drives the backend with. More...

#include <conformance.hpp>

Public Attributes
std::string	model_id = "conformance/model"
std::string	prompt = "You are a careful assistant. Answer briefly. Question: what is a prefix cache?"
std::int32_t	prefix_len = -1
std::int32_t	max_tokens = 16
bool	expect_deterministic = true
	Require identical output for identical cold inputs at temperature 0.
std::string	tensor_op = "identity"
	Tensor-path op; the backend must return its single input unchanged.

Detailed Description

Inputs the kit drives the backend with.

Member Data Documentation

expect_deterministic

bool continuum::backend::ConformanceOptions::expect_deterministic = true

Require identical output for identical cold inputs at temperature 0.

max_tokens

std::int32_t continuum::backend::ConformanceOptions::max_tokens = 16

model_id

std::string continuum::backend::ConformanceOptions::model_id = "conformance/model"

prefix_len

std::int32_t continuum::backend::ConformanceOptions::prefix_len = -1

prompt

std::string continuum::backend::ConformanceOptions::prompt = "You are a careful assistant. Answer briefly. Question: what is a prefix cache?"

Token-path prompt. Its first prefix_len bytes act as the cached prefix on the warm run (default: half the prompt).

tensor_op

std::string continuum::backend::ConformanceOptions::tensor_op = "identity"

Tensor-path op; the backend must return its single input unchanged.

---

The documentation for this struct was generated from the following file:

• include/continuum/backend/conformance.hpp