This commit is contained in:
XXXXRT666 2025-10-24 05:00:54 +01:00
parent 0969145073
commit 99fba5154e
3 changed files with 15 additions and 10 deletions

View File

@ -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

View File

@ -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")

View File

@ -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 |
</div>