search

mudler / LocalAI

:robot: The free, Open Source OpenAI alternative. Self-hosted, community-driven and local-first. Drop-in replacement for OpenAI running on consumer-grade hardware. No GPU required. Runs gguf, transformers, diffusers and many more models architectures. It allows to generate Text, Audio, Video, Images. Also with voice cloning capabilities.
https://localai.io
MIT License
21.75k stars 1.66k forks source link

Vall-E-X backend still broken #1394

Closed TwinFinz closed 6 months ago

TwinFinz commented 7 months ago

LocalAI version: v2.0.0-cublas-cuda12-ffmpeg

Environment, CPU architecture, OS, and Version: Linux LocalAi-GPT 5.15.133.1-microsoft-standard-WSL2 #1 SMP Thu Oct 5 21:02:42 UTC 2023 x86_64 GNU/Linux

Describe the bug Fails to find something (expect it is tokenizer its failing to find)

To Reproduce curl http://localhost:8080/tts -H "Content-Type: application/json" -H "Authorization: Bearer sk-xxx" -d '{ "backend": "bark", "input":"Hello!" }'

Expected behavior Output Audio

Logs 1:31AM DBG GRPC(-127.0.0.1:34935): stderr Traceback (most recent call last): 1:31AM DBG GRPC(-127.0.0.1:34935): stderr File "/build/backend/python/vall-e-x/ttsvalle.py", line 14, in 1:31AM DBG GRPC(-127.0.0.1:34935): stderr from utils.generation import SAMPLE_RATE, generate_audio, preload_models 1:31AM DBG GRPC(-127.0.0.1:34935): stderr File "/build/backend/python/vall-e-x/utils/generation.py", line 47, in 1:31AM DBG GRPC(-127.0.0.1:34935): stderr text_tokenizer = PhonemeBpeTokenizer(tokenizer_path="./utils/g2p/bpe_69.json") 1:31AM DBG GRPC(-127.0.0.1:34935): stderr File "/build/backend/python/vall-e-x/utils/g2p/init.py", line 13, in init 1:31AM DBG GRPC(-127.0.0.1:34935): stderr self.tokenizer = Tokenizer.from_file(tokenizer_path) 1:31AM DBG GRPC(-127.0.0.1:34935): stderr Exception: No such file or directory (os error 2)

Additional context this is a fresh pull of the LocalAi v2 image docker was freshly reinstalled/reset with all previous images removed prior.

mudler commented 7 months ago

Vall-e-x is currently fixed in master. Closing the issue

bdqfork commented 6 months ago

I got same problem today, used docker image quay.io/go-skynet/local-ai:v2.3.0-cublas-cuda12-ffmpeg

mudler commented 6 months ago

I've tested it in the container today and could confirm - the fix was easy, the problem was that unfortunately tests don't cover that e2e. It should be fixed in master now!