SAVRN
Search Contact SAVRN

Open-weight model · Speech recognition

wav2vec2-large-xlsr-mvc-swahili

by Eddie Gulay eddiegulay/wav2vec2-large-xlsr-mvc-swahili

This model is a finetuned version of facebook/wav2vec2-large-xlsr-53. There was an issue with vocab, seems like there are special characters included and they were not considered during training You could try

Parameters315M
Context—
Weights1.3 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads771.2k

Runs On

What it takes to serve wav2vec2-large-xlsr-mvc-swahili (315M parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 0.6 GB 0.8 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 0.3 GB 0.4 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 0.2 GB 0.2 GB 1x MI300X (192 GB)
Vultr		 $1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

SAVRN's Notes on wav2vec2-large-xlsr-mvc-swahili

Swahili transcription is where Eddie Gulay pointed this fine-tune of facebook/wav2vec2-large-xlsr-53, a 315M-parameter CTC model with a 59-entry vocabulary trained on Common Voice 13.0. Memory is not the constraint: 0.6 GB of weights and 0.8 GB needed at 16-bit. The cheapest setup on our list is a single MI300X with 192 GB at $1.85 per hour on-demand, leaving almost the whole card free; the question is what else shares it.

Apache 2.0 permits commercial use, modification and redistribution; keep the license and copyright notices with the files and state significant changes. Check two things. First, the only reported evaluation is the publisher's own: a word error rate of 0.2 on the Swahili split of Common Voice 13.0. Second, the publisher flags a vocabulary problem, special characters not handled during training, and the repository has not been updated since 2023-11-09. Test on your own Swahili audio first.

SAVRN Research, 2026-09-18

Model Card

By Eddie Gulay, published under apache-2.0, revision fc0d824e4eba.

This model is a finetuned version of facebook/wav2vec2-large-xlsr-53. There was an issue with vocab, seems like there are special characters included and they were not considered during training You could try

Read Eddie Gulay's full model card

This model is a finetuned version of facebook/wav2vec2-large-xlsr-53.

How to use the model

There was an issue with vocab, seems like there are special characters included and they were not considered during training
You could try 

from transformers import AutoProcessor, AutoModelForCTC

repo_name = "eddiegulay/wav2vec2-large-xlsr-mvc-swahili"
processor = AutoProcessor.from_pretrained(repo_name)
model = AutoModelForCTC.from_pretrained(repo_name)

# if you have GPU
# move model to CUDA
model = model.to("cuda")


def transcribe(audio_path):
  # Load the audio file
  audio_input, sample_rate = torchaudio.load(audio_path)
  target_sample_rate = 16000
  audio_input = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=target_sample_rate)(audio_input)

  # Preprocess the audio data
  input_dict = processor(audio_input[0], return_tensors="pt", padding=True, sampling_rate=16000)

  # Perform inference and transcribe
  logits = model(input_dict.input_values.to("cuda")).logits
  pred_ids = torch.argmax(logits, dim=-1)[0]
  transcription = processor.decode(pred_ids)

  return transcription

transcript = transcribe('your_audio.mp3')

Configuration

Architecture
Wav2Vec2ForCTC
Layers
24
Hidden size
1,024
Feed-forward size
4,096
Attention heads
16
Vocabulary size
59
Stored precision
float32
Model type
wav2vec2

Identity and Version

Repository
eddiegulay/wav2vec2-large-xlsr-mvc-swahili
Publisher
Eddie Gulay
Task
Speech recognition
Modality
Audio
Library
transformers
Parameters
315M parameters
Languages
sw
Revision
fc0d824e4ebacc09562436993497d350e9ac97c3
First published
2023-11-06
Last updated
2023-11-09

Files and Weights

14 files, 1.3 GB in total. The weights are 2 files totalling 1.3 GB in bin, safetensors.

Weights2 files · 1.3 GB
Configuration5 files · 4.0 KB
Tokenizer2 files · 1.8 KB
Documentation1 file · 1.9 KB
Other3 files · 23.5 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
model.safetensorsWeights1.3 GB 85f0fc528185
training_args.binWeights4.6 KB 0af7d2429472
added_tokens.jsonConfiguration30 B —
config.jsonConfiguration2.3 KB —
inference.pyConfiguration887 B —
preprocessor_config.jsonConfiguration214 B —
special_tokens_map.jsonConfiguration519 B —
README.mdDocumentation1.9 KB —
runs/Nov06_21-17-21_8d905d1e3af6/events.out.tfevents.1699306553.8d905d1e3af6.1183.0Other7.3 KB 243673dcb000
runs/Nov06_22-52-07_8d905d1e3af6/events.out.tfevents.1699311700.8d905d1e3af6.1183.1Other6.1 KB 51d8f6586636
runs/Nov06_23-24-19_8d905d1e3af6/events.out.tfevents.1699313622.8d905d1e3af6.3490.0Other10.1 KB dbad0e156e41
.gitattributesRepository1.5 KB —
tokenizer_config.jsonTokenizer1.2 KB —
vocab.jsonTokenizer657 B —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
1.3 GB
Download from Eddie Gulay

Released by Eddie Gulay through its official repository on Hugging Face. Read the license.

Built From

  • Derived from facebook/wav2vec2-large-xlsr-53
  • Trained on (disclosed) common_voice_13_0

Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

BenchmarkConditionsResultReported byRevisionDate
common_voice_13_0 Configuration swTask Automatic Speech RecognitionMetric WerComparison conditions not established 0.2 eddiegulay
Publisher reported		 Evaluated revision not stated —

Memory Requirements

PrecisionWeights in memory
As published1.3 GB
16-bit0.6 GB
8-bit0.3 GB
4-bit0.2 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About wav2vec2-large-xlsr-mvc-swahili

How much GPU memory does wav2vec2-large-xlsr-mvc-swahili need?

About 0.8 GB at 16-bit and 0.2 GB at 4-bit: the weights (315M parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run wav2vec2-large-xlsr-mvc-swahili on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

Can I use wav2vec2-large-xlsr-mvc-swahili commercially?

Yes. wav2vec2-large-xlsr-mvc-swahili is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Similar Models

Model · Speech recognition

wav2vec2-large-xls-r-300m-Urdu

Abid Ali Awan

A fine-tuned XLS-R 300M CTC model for Urdu automatic speech recognition. It transcribes 16 kHz mono audio and includes an optional 5-gram KenLM decoder. Best reported result: 39.89% WER / 16.70% CER with KenLM decoding on the Urdu Common Voice 8.0 test set. See the Kaggle evaluation notebook for a reproducible example. The repository contains a 5-gram KenLM language model. The Kaggle notebook evaluates a five-sample streaming smoke test from fixie-ai/commonvoice170 (ur, test). Results are reported on the Urdu test split of Mozilla Common Voice 8.0. The language-model row is the model-card score; compare each result only with the same decoding strategy. To reproduce language-model evaluation…

Open weights apache-2.0 315M parameters transformers
View model

Model · Speech recognition

Wav2Vec2-large-xlsr-hindi

Shyam Sunder Kumar

Fine-tuned facebook/wav2vec2-large-xlsr-53 hindi using the Multilingual and code-switching ASR challenges for low resource Indian languages. When using this model, make sure that your speech input is sampled at 16kHz. The model can be used directly (without a language model) as follows: The model can be evaluated as follows on the hindi test data of Common Voice. The script used for training can be found Hindi ASR Fine Tuning Wav2Vec2

Open weights 316M parameters transformers
View model

Model · Speech recognition

koelsch-wav2vec2-ipa

Chem

IPA phoneme recognition for Kölsch (Ripuarian German, Cologne). Fine-tuned from facebook/wav2vec2-xls-r-300m with a CTC head over a 48-symbol IPA inventory. This is also the forced-alignment model. Its frame-level CTC posteriors drive torchaudio.functional.forcedalign to produce Praat TextGrids with word and phone tiers — see notebook 9 in the repository. Trained during the CIF Tandem Fellowship at IfL-Phonetik, University of Cologne, on Alles Kölsch (Bhatt & Lindlar 1998) — 4,670 utterances, 4.5 hours of spontaneous Cologne dialect speech from 105 speakers aged 10–88 across 49 neighbourhoods. Kölsch is Ripuarian German. It has no public speech dataset and no standardised spelling: 94.1 %…

Open weights cc-by-nc-sa-4.0 315M parameters transformers
View model

This model for Croatian ASR is based on the facebook/wav2vec2-xls-r-300m model and was fine-tuned with 300 hours of recordings and transcripts from the ASR Croatian parliament dataset ParlaSpeech-HR v1.0. If you use this model, please cite the following paper: Nikola Ljubešić, Danijel Koržinek, Peter Rupnik, Ivo-Pavao Jazbec. ParlaSpeech-HR -- a freely available ASR dataset for Croatian bootstrapped from the ParlaMint corpus. http://www.lrec-conf.org/proceedings/lrec2022/workshops/ParlaCLARINIII/pdf/2022.parlaclariniii-1.16.pdf Evaluation is performed on the dev and test portions of the ParlaSpeech-HR v1.0 dataset. There are multiple models available, and in terms of CER and WER, the…

Open weights 315M parameters transformers
View model

Finetuned version of KBs VoxRex large model using Swedish radio broadcasts, NST and Common Voice data. Evalutation without a language model gives the following: WER for NST + Common Voice test set (2% of total sentences) is 2.5%. WER for Common Voice test set is 8.49% directly and 7.37% with a 4-gram language model. When using this model, make sure that your speech input is sampled at 16kHz. This model has been fine-tuned for 120000 updates on NST + CommonVoice and then for an additional 20000 updates on CommonVoice only. The additional fine-tuning on CommonVoice hurts performance on the NST+CommonVoice test set somewhat and, unsurprisingly, improves it on the CommonVoice test set. It seems…

Open weights cc0-1.0 315M parameters transformers
View model

Model · Speech recognition

wav2vec2-xls-r-300m-cs-250

Vojtěch Drábek

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the commonvoice 8.0 dataset as well as other datasets listed below. It achieves the following results on the evaluation set: The eval.py script results using a LM are: Fine-tuned facebook/wav2vec2-large-xlsr-53 on Czech using the Common Voice dataset. When using this model, make sure that your speech input is sampled at 16kHz. The model can be used directly (without a language model) as follows: The model can be evaluated using the attached eval.py script: The Common Voice 8.0 train and validation datasets were used for training, as well as the following datasets: - Šmídl, Luboš and Pražák, Aleš, 2013, OVM – Otázky…

Open weights apache-2.0 315M parameters transformers
View model