Model reference · open weights
kosp2e-asr-ko is an open-weight audio or speech model from espnet. kosp2e-asr-ko (BF16) weighs 447 MB; the smallest configuration that runs it is RTX 3060 12 GB.
Summary of the espnet/kosp2e-asr-ko model card, 2026-10-01
What it is
| Released by | espnet |
|---|---|
| Released | 2025-11-21 |
| VRAM | 447 MB for the weights |
What it runs on
| Card | One stream | Memory |
|---|---|---|
| RTX 3060 12 GB | fits | 11.6 GB |
| RTX 4060 Ti 16 GB | fits | 15.4 GB |
| RTX 3090 24 GB | fits | 23.4 GB |
| RTX 4090 24 GB | fits | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
This is the ESPnet2 recipe for the KoSP2E (Korean Speech Perception and Production Experiment) dataset.
The KoSP2E dataset is a large-scale Korean speech corpus designed for speech perception and production experiments. This recipe provides a full ASR pipeline using ESPnet2 with both Transformer and Conformer architectures.
Environment
| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| test | 2320 | 22337 | 77.1 | 20.4 | 2.6 | 4.4 | 27.4 | 76.4 |
| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| test | 2320 | 84267 | 92.5 | 5.7 | 1.8 | 1.7 | 9.2 | 76.4 |
| dataset | Snt | Wrd | Corr | Sub | Del | Ins | Err | S.Err |
|---|---|---|---|---|---|---|---|---|
| test | 2320 | 65361 | 89.4 | 8.6 | 2.0 | 2.1 | 12.7 | 76.4 |
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
How it works