diff --git a/docs/source/features/speculative_decoding/eagle/audio_tts_eagle.md b/docs/source/features/speculative_decoding/eagle/audio_tts_eagle.md index 90558a8f..af0ff6ef 100644 --- a/docs/source/features/speculative_decoding/eagle/audio_tts_eagle.md +++ b/docs/source/features/speculative_decoding/eagle/audio_tts_eagle.md @@ -82,6 +82,7 @@ python3 tools/spec_benchmark.py \ - `--base-model-path`: 基础模型路径(必需) - `--eagle-model-path`: Eagle辅助模型路径(必需) - `--model-id`: 模型标识符(必需) +- `--is-tts`: 是否为TTS模型,默认为False **基准测试配置:** - `--bench-name`: 基准数据集名称,可参考【`tts_fake_data`】 @@ -121,7 +122,9 @@ python3 tools/spec_benchmark.py \ --model-id cosyvoice3 \ --mode both \ --output-dir ./results \ - --deploy-backend pytorch_tts \ + --deploy-backend pytorch \ + --is-tts \ + --bench-name tts_fake_data \ --generate-audio ``` @@ -146,5 +149,7 @@ python3 tools/spec_benchmark.py \ --model-id cosyvoice3 \ --mode both \ --output-dir ./results \ - --deploy-backend pytorch_tts + --deploy-backend pytorch \ + --is-tts \ + --bench-name tts_fake_data \ ``` \ No newline at end of file