Downloads · 30 days
2
1% of all-time downloads
visualears-xlsr-300m-v2-cont-step1568-fp16
Reza2kn/visualears-xlsr-300m-v2-cont-step1568-fp16
visualears-xlsr-300m-v2-cont-step1568-fp16 is a automatic speech recognition model from Reza2kn. Use it when you need speech turned into text. The card lists the license as apache-2.0.
This is the FP16 export of Reza2kn/visualears-xlsr-300m-v2-cont-step1568 for the VisualEars Persian ASR relabeling pipeline.
All-time downloads
186
Public
Parameters
316M
631 MB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors631 MB · 100%
At a glance
- Task
- Automatic Speech Recognition
- License
- apache-2.0
- Model type
- wav2vec2
- Access
- Public
- Created
- Jun 8, 2026
- Updated
- Jun 8, 2026
- SHA
- 38fa5570
Base models
Model card
From the Hugging Face model README
VisualEars XLS-R 300M v2 Step1568 FP16
This is the FP16 export of
Reza2kn/visualears-xlsr-300m-v2-cont-step1568
for the VisualEars Persian ASR relabeling pipeline.
The FP16 path was selected after parity checks against FP32. Lower precision compressed variants were rejected for relabeling.
Precision Decision
Accepted:
float16
Rejected for relabeling:
NVFP4NVFP4A16FP8FP8_DYNAMIC
FP16 Parity
269-row validation parity against FP32:
- Char similarity vs FP32:
0.9995259540175397 - CER vs FP32:
0.0009786651008025053 - WER vs FP32:
0.005176876617773943
Golha v2 benchmark:
| Mode | WER | CER | Mean decode |
|---|---|---|---|
| FP32 | 25.1066% | 7.6194% | 98.1 ms |
| FP16 | 25.0457% | 7.6194% | 31.8 ms |
Golha v2 FP16-vs-FP32:
- Char similarity vs FP32:
0.9994312796208531 - Changed normalized rows:
3 / 69
Usage
import torch
from transformers import AutoModelForCTC, AutoProcessor
model_id = "Reza2kn/visualears-xlsr-300m-v2-cont-step1568-fp16"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForCTC.from_pretrained(model_id, dtype=torch.float16).cuda().eval()
Snapshot State
Pipeline state, parity logs, and Golha v2 comparison artifacts are archived at:
Reza2kn/visualears-full-pipeline-state
Snapshot path:
snapshots/2026-06-08-fp16-ready/
- Task
- Automatic Speech Recognition
- Type
- wav2vec2
- License
- apache-2.0
- Languages
- fa
- Created
- Jun 8, 2026
- Updated
- Jun 8, 2026