Skip to content

Commit b00ddc2

Browse files
Update concept-custom-neural.md
1 parent ed4ade6 commit b00ddc2

File tree

1 file changed

+7
-4
lines changed

1 file changed

+7
-4
lines changed

articles/ai-services/document-intelligence/concept-custom-neural.md

Lines changed: 7 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -276,8 +276,8 @@ https://{endpoint}/formrecognizer/documentModels/{modelId}:copyTo?api-version=20
276276
:::moniker range="doc-intel-4.0.0"
277277

278278
## Billing
279-
280-
Starting with version `2024-07-31-preview` and later you can receive **10 hours** of free model training. Billing charges are calculated for model trainings that exceed 10 hours. You can choose to spend all of 10 free hours on a single build with a large set of data, or utilize it across multiple builds by adjusting the maximum duration value for the `build` operation by specifying `maxTrainingHours`:
279+
280+
Starting with version `2024-07-31-preview`, you can train your custom neural model for longer durations than 30 minutes. Previous versions have been capped at 30 minutes per training instance, with a total of 20 free training instances per month. Now with `2024-07-31-preview`, you can receive **10 hours** of free model training, and train a model for as long as 10 hours. If you would like to train a model for longer than 10 hours, billing charges are calculated for model trainings that exceed 10 hours. You can choose to spend all of 10 free hours on a single build with a large set of data, or utilize it across multiple builds by adjusting the maximum duration value for the `build` operation by specifying `maxTrainingHours` as below:
281281

282282
```bash
283283

@@ -288,7 +288,10 @@ POST /documentModels:build
288288
}
289289
```
290290

291-
Build time varies. Billing is calculated for the actual time spent (excluding time in queue), with a minimum of 30 minutes per training job. The elapsed time is converted to V100 equivalent training hours and reported as part of the model.
291+
> [!NOTE]
292+
> For versions `v3.1 (2023-07031)` and `v3.0 (2022-08-31)`, custom neural model's paid training is not enabled. For the two versions, you will get a maximum of 30 minutes training duration per model.
293+
294+
Each training hour is the amount of compute a single V100 GPU can perform in an hour. As each build takes different amount of time, billing is calculated for the actual time spent (excluding time in queue), with a minimum of 30 minutes per training job. The elapsed time is converted to V100 equivalent training hours and reported as part of the model.
292295

293296
```bash
294297

@@ -311,4 +314,4 @@ Learn to create and compose custom models:
311314
312315
> [!div class="nextstepaction"]
313316
> [**Build a custom model**](how-to-guides/build-a-custom-model.md)
314-
> [**Compose custom models**](how-to-guides/compose-custom-models.md)
317+
> [**Compose custom models**](how-to-guides/compose-custom-models.md)

0 commit comments

Comments
 (0)