KasuleTrevor

cdli-qwen3-asr-en-uganda-severity-disorder_5

Deploy Dedicated

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Get help setting up a custom Dedicated Endpoints.

Talk with our engineer to get a quote for reserved GPU instances with discounts.

README

License: apache-2.0

Evaluation Results

Overall Metrics

Corpus WER (normalized): 0.2293
Corpus CER (normalized): 0.1391
Avg utterance WER (normalized): 0.2229
Avg utterance CER (normalized): 0.1393

Severity Breakdown

Table
severity_speech_impairment	n_samples	n_speakers	avg_wer	avg_cer	median_wer	median_cer	corpus_wer_normalized	corpus_cer_normalized
Severe (frequent breakdowns)	339	3	0.240114	0.162966	0.2	0.123077	0.238321	0.158928
Moderate (requires effort to understand)	340	3	0.230775	0.140405	0.2	0.117647	0.223923	0.134637
Mild (easily understood with minimal effort)	334	3	0.197309	0.114076	0.169048	0.0877193	0.225444	0.125203

Speaker Breakdown

Table
speaker_id	n_samples	n_speakers	avg_wer	avg_cer	median_wer	median_cer	corpus_wer_normalized	corpus_cer_normalized
UG052	112	1	0.289668	0.184329	0.256579	0.165423	0.303597	0.190225
UG042	92	1	0.287833	0.198686	0.235294	0.159365	0.259789	0.179225
UG057	96	1	0.280312	0.193209	0.256112	0.174167	0.304406	0.209424
UG022	99	1	0.247037	0.12257	0.222222	0.104418	0.257774	0.124656
UG001	102	1	0.240097	0.155166	0.195238	0.113134	0.248204	0.158857
UG036	151	1	0.221817	0.133227	0.185185	0.10687	0.215533	0.129047
UG058	151	1	0.185483	0.121975	0.147059	0.0946746	0.183212	0.117258
UG021	77	1	0.162678	0.0905925	0.142857	0.0815217	0.165541	0.0931486
UG014	133	1	0.127478	0.0762418	0.0909091	0.0434783	0.131197	0.0750139

Model provider

KasuleTrevor

Model tree

Base

Qwen/Qwen3-ASR-1.7B

Fine-tuned

this model

Modalities

Input

Audio

Output

Text

Pricing

Dedicated Endpoints

Supported Functionality

Model APIs

Dedicated Endpoints

Container

More information

Explore FriendliAI today

Get started Talk to an engineer