Model
German Citrinet 1024 model.
Use the NGC CLI to download:
Copied!
4 Versions
1.5.0Selected11/18/2021 10:22 PM UTC507.64 MB400 Epochs Copied!
1.5.0Selected
11/18/2021 10:22 PM UTC507.64 MB400 Epochs
Copied!
ACCURACY BEAM SEARCH SIZE 128
| Key | Value |
|---|---|
| MCV7.0 DEV | 5.00 |
| MCV7.0 TEST | 5.84 |
| MLS DEV | 3.76 |
| MLS TEST | 4.5 |
| VOX POPULI DEV | 11.79 |
| VOX POPULI TEST | 9.42 |
ACCURACY GREEDY
| Key | Value |
|---|---|
| MCV7.0 DEV | 6.63 |
| MCV7.0 TEST | 7.59 |
| MLS DEV | 4.06 |
| MLS TEST | 5.07 |
| VOX POPULI DEV | 12.33 |
| VOX POPULI TEST | 10.02 |
MODEL
| Key | Value |
|---|---|
| ARCHITECTURE | CITRINET |
| INPUT | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| OUTPUTS | TRANSCRIBED SPEECH |
09/21/2021 11:14 PM UTC1.68 GB400 EpochsGPU: V100 Copied!
09/21/2021 11:14 PM UTC1.68 GB400 EpochsGPU: V100
Copied!
ACCURACY BEAM SEARCH SIZE 128
| Key | Value |
|---|---|
| MCV7.0 DEV | 5.39 |
| MCV7.0 TEST | 6.31 |
| MLS DEV | 3.72 |
| MLS TEST | 4.6 |
| VOX POPULI DEV | 12 |
| VOX POPULI TEST | 9.7 |
ACCURACY GREEDY
| Key | Value |
|---|---|
| MCV7.0 DEV | 7.37 |
| MCV7.0 TEST | 8.45 |
| MLS DEV | 4.21 |
| MLS TEST | 5.22 |
| VOX POPULI DEV | 12.96 |
| VOX POPULI TEST | 10.41 |
MODEL
| Key | Value |
|---|---|
| ARCHITECTURE | CITRINET |
| INPUTS | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| OUTPUTS | TRANSCRIBED SPEECH |
09/21/2021 11:13 PM UTC0 B500 Epochs No files available
09/21/2021 11:13 PM UTC0 B500 Epochs
ACCURACY
| Key | Value |
|---|---|
| MCV DEV | 8.36 |
| MCV TEST | 9.55 |
| MLS DEV | 4.82 |
| MLS TEST | 5.93 |
| VOX POPULI DEV | 13.17 |
| VOX POPULI TEST | 10.79 |
MODEL
| Key | Value |
|---|---|
| ARCHITECTURE | CITRINET |
| INPUTS | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| OUTPUTS | TRANSCRIBED SPEECH |
07/27/2021 8:50 PM UTC462.36 MB400 Epochs Copied!
07/27/2021 8:50 PM UTC462.36 MB400 Epochs
Copied!
ACCURACY
| Key | Value |
|---|---|
| MCV dev | 8.1 % |
| MCV test | 9.3 % |
| MLS dev | 5 % |
| MLS test | 6.2 % |
MODEL
| Key | Value |
|---|---|
| ARCHITECTURE | CITRINET |
| INPUTS | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| OUTPUTS | TRANSCRIBED SPEECH |