From 99fba5154ea9483457fede37f9216e6ac94842d9 Mon Sep 17 00:00:00 2001 From: XXXXRT666 <157766680+XXXXRT666@users.noreply.github.com> Date: Fri, 24 Oct 2025 05:00:54 +0100 Subject: [PATCH] . --- GPT_SoVITS/Accelerate/PyTorch/t2s_engine.py | 5 +++++ GPT_SoVITS/TTS_infer_pack/TTS.py | 2 +- README.md | 18 +++++++++--------- 3 files changed, 15 insertions(+), 10 deletions(-) diff --git a/GPT_SoVITS/Accelerate/PyTorch/t2s_engine.py b/GPT_SoVITS/Accelerate/PyTorch/t2s_engine.py index 6f9319e3..1d348802 100644 --- a/GPT_SoVITS/Accelerate/PyTorch/t2s_engine.py +++ b/GPT_SoVITS/Accelerate/PyTorch/t2s_engine.py @@ -173,6 +173,11 @@ class T2SEngine(T2SEngineProtocol): if idx == 10: torch_profiler.end() + + if debug: + timer.summary() + timer.clear() + finally: if ( request.use_cuda_graph diff --git a/GPT_SoVITS/TTS_infer_pack/TTS.py b/GPT_SoVITS/TTS_infer_pack/TTS.py index 0745d0bc..aaa661d8 100644 --- a/GPT_SoVITS/TTS_infer_pack/TTS.py +++ b/GPT_SoVITS/TTS_infer_pack/TTS.py @@ -1334,7 +1334,7 @@ class TTS: gc.collect() if self.configs.device.type == "cuda": - logger.info(str(torch.cuda.memory_allocated(self.configs.device.index) / 1024**3), "GB") + logger.info(str(torch.cuda.memory_allocated(self.configs.device.index) / 1024**3) + "GB") elif self.configs.device.type == "mps": logger.info("Curr: " + str(torch.mps.current_allocated_memory() / 1024**3) + " GB") logger.info("Driver: " + str(torch.mps.driver_allocated_memory() / 1024**3) + " GB") diff --git a/README.md b/README.md index db6f557a..3b5aa06f 100644 --- a/README.md +++ b/README.md @@ -51,15 +51,15 @@ Unseen speakers few-shot fine-tuning demo: ## Infer Speed -| Device | RTF | TTFT | Batch Size | Backend | -| :---------: | :---: | :----: | :--------: | :-------------------------: | -| RTX 5090 | 0.05 | 150 ms | 1 | Flash Attn Varlen CUDAGraph | -| RTX 4090 | 0.014 | UNK | 24 | Flash Attn Varlen CUDAGraph | -| RTX 4060 Ti | 0.07 | 460 ms | 1 | Flash Attn Varlen CUDAGraph | -| RTX 4060 Ti | 0.028 | UNK | 28 | Flash Attn Varlen CUDAGraph | -| Apple M4 | 0.16 | 1363ms | 1 | MLX Varlen | -| Apple M4 | 0.06 | UNK | 40 | MLX Varlen | -| i7-12700K | 0.28 | UNK | 40 | Torch Varlen | +| Device | RTF | TTFT | Batch Size | Backend | +| :---------: | :----: | :----: | :--------: | :-------------------------: | +| RTX 5090 | 0.05 | 150 ms | 1 | Flash Attn Varlen CUDAGraph | +| RTX 5090 | 0.0122 | UNK | 30 | Flash Attn Varlen CUDAGraph | +| RTX 4060 Ti | 0.07 | 460 ms | 1 | Flash Attn Varlen CUDAGraph | +| RTX 4060 Ti | 0.028 | UNK | 28 | Flash Attn Varlen CUDAGraph | +| Apple M4 | 0.16 | 1363ms | 1 | MLX Varlen | +| Apple M4 | 0.06 | UNK | 40 | MLX Varlen | +| i7-12700K | 0.28 | UNK | 40 | Torch Varlen |