faster-whisper-large-v3-turbo如果只用cpu是能跑起来，但是用gpu的话faster-whisper库要用到ctranslate2才能跑，看了下载资源里最新的只有DAS1.3的（https://download.sourcefind.cn:65024/4/main/ctranslate2/DAS1.3），平时用的镜像版本通常是比较新的比如 harbor.sourcefind.cn:5443/dcu/admin/base/custom:vllm018-ubuntu22.04-dtk26.04-gemma4-0413，请问这种DAS1.3老版本的镜像要去哪里下载呀？ 或者有没有类似模型仓库里那种直接能跑的方案？或者出个新一些版本的ctranslate？ （看之前也有讨论的帖子不知道后续搞定了没 https://forum.sourcefind.cn/thread/1196） faster-whisper库：https://github.com/SYSTRAN/faster-whisper/tree/master 模型：https://www.modelscope.cn/models/pengzhendong/faster-whisper-large-v3-turbo 注：直接在新的镜像比如vllm018里卸载ctranslate2再装DAS1.3的ctranslate2不可行，报错： import ctranslate2 Traceback (most recent call last): File "", line 1, in File "/usr/local/lib/python3.10/dist-packages/ctranslate2/__init__.py", line 21, in from ctranslate2._ext import ( ImportError: /usr/local/lib/python3.10/dist-packages/ctranslate2/libcublas.so.11: undefined symbol: rocblas_ztrmm_outofplace 测试用例是： from faster_whisper import WhisperModel model_dir = "/workspace/models/faster-whisper-large-v3-turbo" model = WhisperModel(model_dir, device="cuda", compute_type="auto", local_files_only=True) segments, info = model.transcribe("dataset/speech.wav") print("Detected language '%s' with probability %f" % (info.language, info.language_probability)) for segment in segments: print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))