Automatic Speech Recognition
Safetensors
Italian
whisper
italian
localai
mudler commited on
Commit
e0b6f28
·
verified ·
1 Parent(s): 60738d3

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +65 -0
README.md ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: it
3
+ license: mit
4
+ tags:
5
+ - whisper
6
+ - automatic-speech-recognition
7
+ - italian
8
+ - localai
9
+ datasets:
10
+ - mozilla-foundation/common_voice_25_0
11
+ - facebook/multilingual_librispeech
12
+ - facebook/voxpopuli
13
+ base_model: openai/whisper-tiny
14
+ pipeline_tag: automatic-speech-recognition
15
+ ---
16
+
17
+ # whisper-tiny-it-multi
18
+
19
+ Fine-tuned [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) (39M params) for Italian ASR on multiple datasets.
20
+
21
+ **Author:** Ettore Di Giacinto
22
+
23
+ Brought to you by the [LocalAI](https://github.com/mudler/LocalAI) team. This model can be used directly with [LocalAI](https://localai.io).
24
+
25
+ ## Results
26
+
27
+ Evaluated on combined test set (Common Voice + MLS + VoxPopuli, 17,598 samples):
28
+
29
+ | Step | WER |
30
+ |------|-----|
31
+ | 1000 | 39.8% |
32
+ | 3000 | 33.5% |
33
+ | 5000 | 31.4% |
34
+ | 10000 | **29.4%** |
35
+
36
+ ## Training Details
37
+
38
+ - **Base model:** openai/whisper-tiny (39M parameters)
39
+ - **Datasets:** Common Voice 25.0 Italian (173k) + MLS Italian (60k) + VoxPopuli Italian (23k) = 255k train samples
40
+ - **Steps:** 10,000 (batch size 32)
41
+ - **Learning rate:** 1e-5 with 500 warmup steps
42
+ - **Precision:** bf16 on NVIDIA GB10
43
+
44
+ ## Usage
45
+
46
+ ### Transformers
47
+
48
+ ```python
49
+ from transformers import pipeline
50
+
51
+ pipe = pipeline("automatic-speech-recognition", model="LocalAI-io/whisper-tiny-it-multi")
52
+ result = pipe("audio.mp3", generate_kwargs={"language": "it", "task": "transcribe"})
53
+ print(result["text"])
54
+ ```
55
+
56
+ ### CTranslate2 / faster-whisper
57
+
58
+ For optimized CPU inference: [LocalAI-io/whisper-tiny-it-multi-ct2-int8](https://huggingface.co/LocalAI-io/whisper-tiny-it-multi-ct2-int8)
59
+
60
+ ## Links
61
+
62
+ - **CV-only version:** [LocalAI-io/whisper-tiny-it](https://huggingface.co/LocalAI-io/whisper-tiny-it) (WER 27.1% on CV test)
63
+ - **CTranslate2 INT8:** [LocalAI-io/whisper-tiny-it-multi-ct2-int8](https://huggingface.co/LocalAI-io/whisper-tiny-it-multi-ct2-int8)
64
+ - **Code:** [github.com/localai-org/whisper-it](https://github.com/localai-org/whisper-it)
65
+ - **LocalAI:** [github.com/mudler/LocalAI](https://github.com/mudler/LocalAI)