File size: 2,305 Bytes
371456c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
---
license: apache-2.0
datasets:
- asierhv/composite_corpus_eu_v2.1
language:
- eu
metrics:
- wer
model-index:
- name: Whisper Base Basque
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 18.0
      type: mozilla-foundation/common_voice_18_0
      config: eu
      split: test
      args:
        language: eu
    metrics:
      - name: Test WER
        type: wer
        value: 10.78
base_model:
- xezpeleta/whisper-base-eu
---

# Whisper Basque (eu) - CTranslate2 Conversion (int8)

**This is a CTranslate2 conversion of [xezpeleta/whisper-base-eu](https://huggingface.co./xezpeleta/whisper-base-eu) designed for use with faster-whisper.**

## Model Details
- **Base Model:** OpenAI Whisper Base (original model card: [whisper-base](https://huggingface.co./openai/whisper-base))
- **Finetuned for:** Basque (eu) speech recognition
- **Dataset:** `asierhv/composite_corpus_eu_v2.1` (Mozilla Common Voice 18.0 + Basque Parliament + OpenSLR)
- **Conversion Format:** CTranslate2 (optimized for inference)
- **Compatibility:** Designed for use with [faster-whisper](https://github.com/SYSTRAN/faster-whisper)
- **Quantization:** int8 (ready for CPU inference)
- **WER:** 10.78% on Mozilla Common Voice 18.0

## Usage with faster-whisper

First install required packages:
```bash
pip install faster-whisper
```

Then use the following code snippet:

```py
from faster_whisper import WhisperModel

# Load the model (FP16 precision)
model = WhisperModel("xezpeleta/whisper-base-eu-ct2", device="cuda", compute_type="float16")

# Transcribe audio file
segments, info = model.transcribe("audio.mp3", language="eu")

# Print transcription
for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))
```

## Evaluation

The model achieves **10.78% Word Error Rate (WER)** on the Basque `test` split of **Mozilla Common Voice 18.0**.


# Conversion details

Converted from the original HuggingFace model using:

```bash
ct2-transformers-converter --model xezpeleta/whisper-base-eu \
                           --output_dir whisper-base-eu-ct2 \
                           --copy_files tokenizer.json preprocessor_config.json \
                           --quantization float16
```