veelion comments

Results 9 comments of


                                            veelion

Undefined reference when compile decode_main with GPU support

Maybe the cuda version build with libtorch_cuda_cpp.so, does not match the cuda version installed on the system.

OPUS support problems

> That's entirely up to https://github.com/libsndfile/libsndfile. libsndfile has supported MP3 and OPUS, any plan to python-soundfile? I tried to use soundfile with libsndfile(1.0.31) to read MP3 file successfully, but failed...

What is the Max number of translation?

I'm not very clear with Google's blocking policy. But I found something: I was blocked by Google many times, if requests with different texts again after a few seconds, the...

Plain pytorch LLaMA implementation (no fairscale, use as many GPUs as you want)

Great work！ But I got an error: NotImplementedError: Cannot copy out of meta tensor; no data!, which comes from `accelerate` ``` ╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮ │ ~/github/vanilla-llama/example.py:18...

Plain pytorch LLaMA implementation (no fairscale, use as many GPUs as you want)

Run 7B model on 1 GPU (1070, 8GB) ``` python example.py --llama-path models/ --model 7B ```

Plain pytorch LLaMA implementation (no fairscale, use as many GPUs as you want)

Hi @galatolofederico , I have tried convert the weights many times, but still got the same error as before. I followed @randaller 's method to print device_map, all are on...

support silence mode

同关注这个问题。试了“嗯嗯、嗯嗯”这样的语音，不能生成闭嘴视频

support silence mode

> 如何支持静音闭嘴功能您有解决办法了吗？

BUG 多个请求qwen2-7b模型时，推理会报错 probability tensor contains either `inf`, `nan` or element < 0

通过 dify_client 调用 API时，遇到同样的错误，然而在dify自己的聊天界面使用时就没有这个错误。