Apr 7

1. 08:20

   Restore full ASR model - all fixes applied

   Tsedee
   Verified signature

   6c09762
2. 07:41

   Restore ASR app with suppress_tokens fix (dev mode disabled)

   Tsedee
   Verified signature

   1a3c6b1
3. 07:28

   Debug: minimal ZeroGPU+Gradio test (no model)

   Tsedee
   Verified signature

   80e1752
4. 07:07

   Fix 500: remove server_name param, keep default launch()

   Tsedee
   Verified signature

   26a4baa
5. 06:34

   Fix 500: low_cpu_mem_usage, server_name=0.0.0.0, remove container=False

   Tsedee
   Verified signature

   628ccba
6. 05:50

   Fix AttributeError: remove generation_config.language from print statement

   Tsedee
   Verified signature

   ca4c3dd
7. 05:44

   Fix IndexError: set suppress_tokens=None (was empty [] in generation_config)

   Tsedee
   Verified signature

   e8bceef
8. 05:39

   Add error catching to display full traceback in UI for debugging

   Tsedee
   Verified signature

   f75ed37
9. 05:36

   Fix ValueError: separate model.half() and model.to(cuda), remove language/task from generate

   Tsedee
   Verified signature

   d8e1caf
10. 05:20

    Fix IndexError: replace forced_decoder_ids with language/task in generate()

    Tsedee
    Verified signature

    f4209ae
11. 05:08

    Fix ZeroGPU: load CPU at startup, move to GPU inside @spaces.GPU only

    Tsedee
    Verified signature

    d2249bd
12. 04:54

    Fix ZeroGPU: load model to GPU at startup, create pipe globally

    Tsedee
    Verified signature

    6852ec9
13. 04:32

    Add spaces library for ZeroGPU support

    Tsedee
    Verified signature

    0160c0c
14. 04:32

    GPU: ZeroGPU A10G + float16 + @spaces.GPU decorator

    Tsedee
    Verified signature

    5f950b9
15. 04:11

    Redesign: elegant dark-gradient UI, mic 5min, file upload 25min

    Tsedee
    Verified signature

    2210691
16. 02:59

    Fix UI: combine mic+upload into single Audio component for Gradio 5.x

    Tsedee
    Verified signature

    3aa0b76
17. 02:44

    Fix: load processor from base model (adapter repo has no tokenizer files)

    Tsedee
    Verified signature

    ea1a9b0
18. 02:37

    Fix: monkey-patch peft _torch_supports_distributed=False to bypass DTensor AttributeError

    Tsedee
    Verified signature

    0e42741
19. 02:29

    Fix: pin peft==0.18.1 and torch>=2.4.0 for DTensor compatibility

    Tsedee
    Verified signature

    f6f6881
20. 02:29

    Fix: pre-import torch.distributed.tensor + pin peft==0.18.1 torch>=2.4.0

    Tsedee
    Verified signature

    ea23fef
21. 02:17

    Fix: Add peft library for LoRA adapter loading

    Tsedee
    Verified signature

    bd33ac7
22. 02:17

    Fix: Load LoRA adapter correctly using PeftModel + base model

    Tsedee
    Verified signature

    3a904dd
23. 01:43

    Fix: Pin tokenizers<0.21 to avoid TemplateProcessing TypeError

    Tsedee
    Verified signature

    3bc299c
24. 01:43

    Fix: Remove WhisperTokenizer, use pipeline generate_kwargs for language/task

    Tsedee
    Verified signature

    c49bfa0
25. 01:20

    Fix: Remove spaces.GPU decorator and torch.cuda.amp.autocast for CPU compatibility

    Tsedee
    Verified signature

    70c664b
26. 00:22

    Fix: remove gradio pin, managed by sdk_version

    Tsedee
    Verified signature

    3dff2b4
27. 00:22

    Fix: set sdk_version to 5.23.0

    Tsedee
    Verified signature

    1e13dbb
28. 00:16

    Fix: remove max_length param (gradio 5 compat)

    Tsedee
    Verified signature

    4f90d18
29. 00:16

    Fix: upgrade to gradio 5.x for huggingface_hub compatibility

    Tsedee
    Verified signature

    acabd22
30. 00:10

    Add README.md

    Tsedee
    Verified signature

    7f5fd3f
31. 00:10

    Add requirements.txt

    Tsedee
    Verified signature

    33f79a0
32. 00:10

    Add app.py

    Tsedee
    Verified signature

    90a4815
33. 00:10

    initial commit

    Tsedee
    Verified signature

    46ddc73