SeedTTSTextTraceFlavorConfig

class SeedTTSTextTraceFlavorConfig ( BaseTraceFlavorConfig )

Seed TTS eval text dataset configuration.

Polymorphic Type:

type: seed_tts_text

All BaseTraceFlavorConfig types:

• timed_synthetic_session: TimedSyntheticSessionTraceFlavorConfig

• shared_prefix: SharedPrefixTraceFlavorConfig

• rag: RAGTraceFlavorConfig

• request_log: RequestLogTraceFlavorConfig

• untimed_content_multi_turn: UntimedContentMultiTurnTraceFlavorConfig

• sharegpt: ShareGPTTraceFlavorConfig

• audio: AudioTraceFlavorConfig

• seed_tts_text: SeedTTSTextTraceFlavorConfig

Fields:

dataset_namestr = "TwinkStart/Seed-TTS-Eval"

Hugging Face dataset name used when local_path is empty.

subsetstr = "en"

Dataset subset/config name. Defaults to English.

splitstr = "train"

Dataset split to load.

text_columnstr = "text"

Column containing the target TTS synthesis text.

id_columnstr = "filename"

Optional source row identifier column copied into request metadata.

local_pathstr = ""

Optional local dataset path. Supports saved HF datasets and common data files such as JSON/JSONL, CSV, and Parquet.

min_tokensint = 20

Minimum input word count. Rows with fewer words are skipped. Ignored when min_chars/max_chars are set.

max_tokensint = 150

Maximum input word count. Text is truncated to a sampled word count in [min_tokens, max_tokens]. Ignored when min_chars/max_chars are set.

min_charsint = -1

If >= 0, enables char-based input length control. Rows with fewer chars are skipped.

max_charsint = -1

If >= 0, enables char-based input length control. Text is truncated to a sampled char count in [min_chars, max_chars].