EvSys

ModelEvalConfig

Attributes

attributemax_tokensint
= 256
attributetemperaturefloat
= 0.0
attributemax_attemptsint
= 5

Retries per generation call (transient inference errors).

attributebatch_sizeint
= 1

When >1 and the client exposes generate_batch, submit prompts in chunks of this size and collect them concurrently. Falls back to sequential generate calls if the client doesn't support batching.

attributeprompt_builderPromptBuilder
= qwen_chat_prompt

Functions

func__init__(self, max_tokens=256, temperature=0.0, max_attempts=5, batch_size=1, prompt_builder=qwen_chat_prompt) -> None
paramself
parammax_tokensint
= 256
paramtemperaturefloat
= 0.0
parammax_attemptsint
= 5
parambatch_sizeint
= 1
paramprompt_builderPromptBuilder
= qwen_chat_prompt

Returns

None

On this page