Examples take audio only (drop empty text instruction column)

Drop conflicting text prompt from translate examples (audio carries its own instruction)

Use two split translate examples

Split translate example into two separate tasks

Point examples at fused audio clips

Fuse example audios per sequence.json into single clips

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Wrap inference in torch.no_grad() to avoid autograd-graph memory pressure

Upload folder using huggingface_hub

Fix kv-cache device mismatch and input_pos accumulation bug

Upload app.py with huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

initial commit