Skip to content
ModelListedExternal
Browse files

Whisper Deepstack Tn Intent

This model is a small classifier head (0.8M params) trained on top of frozen Whisper-large-v3 encoder features for Tunisian Arabic intent recognition as part of the NADI 2026 Shared Task.

Sourcehuggingface.co/FatimahEmadEldin/whisper-deepstack-tn-intent/tree/73eaea2554d082b2ed42010fda36ca797bdc444a
KindOtherVersionBest classifierCgrade

Model source

Kind
Other
File size
3.04–20.4 MB
Version
vBest classifier
Updated
Oct 25, 2026
Source
Hugging Face

Source description

Whisper-DeepStack Tunisian Arabic Intent Classifier

This model is a small classifier head (~0.8M params) trained on top of frozen Whisper-large-v3 encoder features for Tunisian Arabic intent recognition as part of the NADI 2026 Shared Task.

Architecture

  • Encoder: openai/whisper-large-v3 (frozen)
  • DeepStack Layers: [10, 21, 31]
  • Aggregation: sum
  • Validation Accuracy: 0.1513
  • Validation Weighted F1: 0.0882

Usage

import torch
from huggingface_hub import hf_hub_download

ckpt = torch.load(
    hf_hub_download("FatimahEmadEldin/whisper-deepstack-tn-intent", "best_classifier.pt"),
    map_location='cpu'
)

Sources

1 source
Hugging FacePrimaryVersion 73eaea2554d082b2ed42010fda36ca797bdc444a
huggingface.co/FatimahEmadEldin/whisper-deepstack-tn-intent/tree/73eaea2554d082b2ed42010fda36ca797bdc444a

Verified Oct 25

Download files

Model artifacts

2 artifacts

best_classifier.pt

pt · 3.04 MB · SHA-256 1b7107ce9bcc…8963 · Hugging Face

Download file

consolidated_features.pt

pt · 20.4 MB · SHA-256 59d21835e543…a773 · Hugging Face

Download file

Source excerpts

2 excerpts
ReadmeAug 7huggingface.co/FatimahEmadEldin/whisper-deepstack-tn-intent

--- language: ar tags: - audio-classification - intent-classification - arabic - tunisian-arabic - slurp-tn - nadi-2026 license: apache-2.0 datasets: - Elyadata/SLURP-TN --- # Whisper-DeepStack Tunisian Arabic Intent Classifier This model is a small classifier head (~0.8M params) trained on top of frozen Whisper-large-v3 encoder features for Tunisian Arabic intent recognition as part of the NADI 2026 Shared Task. ## Architecture - **Encoder**: `openai/whisper-large-v3` (frozen) - **DeepStack Layers**: [10, 21, 31] - **Aggregation**: sum - **Validation Accuracy**: 0.1513 - **Validation Weighted F1**: 0.0882 ## Usage ```python import torch from huggingface_hub import hf_hub_download ckpt = torch.load( hf_hub_download("FatimahEmadEldin/whisper-deepstack-tn-intent", "best_classifier.pt"), map_location='cpu' ) ```

NoteAug 7huggingface.co/FatimahEmadEldin/whisper-deepstack-tn-intent

Source context: 0 downloads · 0 likes · Pipeline audio-classification · Repo FatimahEmadEldin/whisper-deepstack-tn-intent