add 2

sync local state with remote

remove usage instructions and example text from UI

fix: use torchaudio.info with suppressed deprecation warnings instead of torchcodec

fix: convert non-WAV audio to WAV before encoding, use torchcodec for duration

fix: use torchaudio.info for audio duration (supports m4a without deprecation warnings)

fix: use librosa.get_duration for audio duration check (supports m4a and more formats)

fix: only unpatch ZeroGPU in nanovllm child processes, skip @spaces.GPU workers

fix: forcibly clear ZeroGPU dispatch/function mode stacks in child processes

fix: replace ZeroGPU dispatch stubs with pass-through in child processes instead of trying to pop modes

fix: properly unpatch ZeroGPU dispatch modes and force torch.load map_location=cpu in child processes

fix: exclude assets/ and examples/ from LFS tracking

fix: use functools.partial for pickleable multiprocessing wrapper

fix: make spaces torch mode __exit__ no-op to avoid PyTorch 2.8.0 IndexError

fix: unpatch ZeroGPU torch stubs in every multiprocessing child process

fix: import os before using os.environ in top-level patches

fix: patch spaces ZeroGPU workers to non-daemon to allow child processes

fix: use direct env check to disable prewarm on ZeroGPU

fix: wrap TTS inference in @spaces.GPU for ZeroGPU compatibility

fix: suppress harmless asyncio BaseEventLoop __del__ errors during GC

fix: close asyncio event loop in worker thread to suppress __del__ errors

fix: upgrade torch to 2.8.0 for ZeroGPU compatibility

fix: import spaces before other packages to avoid CUDA init error

fix: harden async nanovllm demo stability

fix: revert concurrency limit to 1 — nanovllm event loop is not thread-safe

fix: remove unsupported inference_timesteps parameter and dit_steps slider

fix: add cookbook guide heading link

fix: store daily request logs on persistent disk

fix: show localized upload alert for long audio

fix: zipenhancer for denoise

fix: add denoiser and reference audio guards

fix: harden Gradio NanoVLLM startup path

fix: install flash-attn from prebuilt wheel

fix: pin Gradio NanoVLLM runtime dependencies

fix: switch Gradio demo backend to nanovllm_voxcpm

add torch version && torch compile == true

add torch version

fix: dynamically install torchaudio matching ZeroGPU torch+CUDA version

fix: install CPU torchaudio to avoid CUDA 13 mismatch on ZeroGPU

fix: install voxcpm with --no-deps at runtime to preserve ZeroGPU torch/torchaudio

fix: remove torch/torchaudio pin, rely on ZeroGPU pre-installed version

fix: pin torch/torchaudio to 2.5.0 and move import voxcpm to module level

fix: move SDPA backend config into GPU function context

fix: disable flash/mem-efficient SDPA backends for ZeroGPU compat

fix: disable optimize to avoid ZeroGPU torch.compile incompatibility

upgrade to VoxCPM2 + Gradio 6 with full i18n UI

Update README.md

set zero gpu inference

set zero gpu inference

set zero gpu inference