Skip to content

Commit 6864cfd

Browse files
prmerger-automator[bot]prmerger-automator[bot]
authored
Merge pull request #159 from mrbullwinkle/mrb_09_05_2024_fine_tuning
[Azure OpenAI] PM Fine-tuning update
2 parents d81c7b9 + e1d49ff commit 6864cfd

File tree

5 files changed

+22
-12
lines changed

5 files changed

+22
-12
lines changed

‎articles/ai-services/openai/how-to/fine-tuning.md

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@ manager: nitinme
77
ms.service: azure-ai-openai
88
ms.custom: build-2023, build-2023-dataai, devx-track-python
99
ms.topic: how-to
10-
ms.date: 08/22/2024
10+
ms.date: 09/05/2024
1111
author: mrbullwinkle
1212
ms.author: mbullwin
1313
zone_pivot_groups: openai-fine-tuning-new

‎articles/ai-services/openai/includes/fine-tuning-openai-in-ai-studio.md

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -32,13 +32,14 @@ The following models support fine-tuning:
3232
- `gpt-35-turbo` (1106)
3333
- `gpt-35-turbo` (0125)
3434
- `gpt-4` (0613)**<sup>*</sup>**
35+
- `gpt-4o` (2024-08-06)**<sup>*</sup>**
3536
- `gpt-4o-mini` (2024-07-18)**<sup>*</sup>**
3637

3738
**<sup>*</sup>** Fine-tuning for this model is currently in public preview.
3839

39-
Consult the [models page](../concepts/models.md#fine-tuning-models) to check which regions currently support fine-tuning.
40+
Or you can fine tune a previously fine-tuned model, formatted as base-model.ft-{jobid}.
4041

41-
If you plan to use `gpt-4` for fine-tuning, please refer to the [GPT-4 public preview safety evaluation guidance](#safety-evaluation-gpt-4-fine-tuning---public-preview)
42+
Consult the [models page](../concepts/models.md#fine-tuning-models) to check which regions currently support fine-tuning.
4243

4344
## Review the workflow for Azure AI Studio
4445

@@ -253,7 +254,7 @@ When each training epoch completes a checkpoint is generated. A checkpoint is a
253254

254255
:::image type="content" source="../media/fine-tuning/checkpoints.png" alt-text="Screenshot of checkpoints UI." lightbox="../media/fine-tuning/checkpoints.png":::
255256

256-
## Safety evaluation GPT-4 fine-tuning - public preview
257+
## Safety evaluation GPT-4, GPT-4o, GPT-4o-mini fine-tuning - public preview
257258

258259
[!INCLUDE [Safety evaluation](../includes/safety-evaluation.md)]
259260

‎articles/ai-services/openai/includes/fine-tuning-python.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -32,13 +32,12 @@ The following models support fine-tuning:
3232
- `gpt-35-turbo` (1106)
3333
- `gpt-35-turbo` (0125)
3434
- `gpt-4` (0613)**<sup>*</sup>**
35+
- `gpt-4o` (2024-08-06)**<sup>*</sup>**
3536
- `gpt-4o-mini` (2024-07-18)**<sup>*</sup>**
3637

3738
**<sup>*</sup>** Fine-tuning for this model is currently in public preview.
3839

39-
If you plan to use `gpt-4` for fine-tuning, please refer to the [GPT-4 public preview safety evaluation guidance](#safety-evaluation-gpt-4-fine-tuning---public-preview)
40-
41-
Or you can fine tune a previously fine-tuned model, formatted as base-model.ft-{jobid}.
40+
Or you can fine tune a previously fine-tuned model, formatted as `base-model.ft-{jobid}`.
4241

4342
:::image type="content" source="../media/fine-tuning/models.png" alt-text="Screenshot of model options with a custom fine-tuned model." lightbox="../media/fine-tuning/models.png":::
4443

@@ -287,6 +286,7 @@ The current supported hyperparameters for fine-tuning are:
287286
|`batch_size` |integer | The batch size to use for training. The batch size is the number of training examples used to train a single forward and backward pass. In general, we've found that larger batch sizes tend to work better for larger datasets. The default value as well as the maximum value for this property are specific to a base model. A larger batch size means that model parameters are updated less frequently, but with lower variance. |
288287
| `learning_rate_multiplier` | number | The learning rate multiplier to use for training. The fine-tuning learning rate is the original learning rate used for pre-training multiplied by this value. Larger learning rates tend to perform better with larger batch sizes. We recommend experimenting with values in the range 0.02 to 0.2 to see what produces the best results. A smaller learning rate can be useful to avoid overfitting. |
289288
|`n_epochs` | integer | The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset. |
289+
|`seed` | integer | The seed controls the reproducibility of the job. Passing in the same seed and job parameters should produce the same results, but may differ in rare cases. If a seed isn't specified, one will be generated for you. |
290290

291291
To set custom hyperparameters with the 1.x version of the OpenAI Python API:
292292

@@ -374,7 +374,7 @@ This command isn't available in the 0.28.1 OpenAI Python library. Upgrade to the
374374

375375
---
376376

377-
## Safety evaluation GPT-4 fine-tuning - public preview
377+
## Safety evaluation GPT-4, GPT-4o, GPT-4o-mini fine-tuning - public preview
378378

379379
[!INCLUDE [Safety evaluation](../includes/safety-evaluation.md)]
380380

‎articles/ai-services/openai/includes/fine-tuning-rest.md

Lines changed: 8 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -31,13 +31,16 @@ The following models support fine-tuning:
3131
- `gpt-35-turbo` (1106)
3232
- `gpt-35-turbo` (0125)
3333
- `gpt-4` (0613)**<sup>*</sup>**
34+
- `gpt-4o` (2024-08-06)**<sup>*</sup>**
3435
- `gpt-4o-mini` (2024-07-18)**<sup>*</sup>**
3536

3637
**<sup>*</sup>** Fine-tuning for this model is currently in public preview.
3738

39+
Or you can fine tune a previously fine-tuned model, formatted as base-model.ft-{jobid}.
40+
3841
Consult the [models page](../concepts/models.md#fine-tuning-models) to check which regions currently support fine-tuning.
3942

40-
If you plan to use `gpt-4` for fine-tuning, please refer to the [GPT-4 public preview safety evaluation guidance](#safety-evaluation-gpt-4-fine-tuning---public-preview).
43+
4144

4245
## Review the workflow for the REST API
4346

@@ -153,6 +156,8 @@ You can create a custom model from one of the following available base models:
153156
- `gpt-35-turbo` (1106)
154157
- `gpt-35-turbo` (0125)
155158
- `gpt-4` (0613)
159+
- `gpt-4o` (2024-08-06)
160+
- `gpt-4o-mini` (2023-07-18)
156161

157162
Or you can fine tune a previously fine-tuned model, formatted as base-model.ft-{jobid}.
158163

@@ -216,6 +221,7 @@ The current supported hyperparameters for fine-tuning are:
216221
|`batch_size` |integer | The batch size to use for training. The batch size is the number of training examples used to train a single forward and backward pass. In general, we've found that larger batch sizes tend to work better for larger datasets. The default value as well as the maximum value for this property are specific to a base model. A larger batch size means that model parameters are updated less frequently, but with lower variance. |
217222
| `learning_rate_multiplier` | number | The learning rate multiplier to use for training. The fine-tuning learning rate is the original learning rate used for pre-training multiplied by this value. Larger learning rates tend to perform better with larger batch sizes. We recommend experimenting with values in the range 0.02 to 0.2 to see what produces the best results. A smaller learning rate can be useful to avoid overfitting. |
218223
|`n_epochs` | integer | The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset. |
224+
|`seed` | integer | The seed controls the reproducibility of the job. Passing in the same seed and job parameters should produce the same results, but may differ in rare cases. If a seed isn't specified, one will be generated for you. |
219225

220226
## Check the status of your customized model
221227

@@ -248,7 +254,7 @@ curl -X POST $AZURE_OPENAI_ENDPOINT/openai/fine_tuning/jobs/{fine_tuning_job_id}
248254
-H "api-key: $AZURE_OPENAI_API_KEY"
249255
```
250256

251-
## Safety evaluation GPT-4 fine-tuning - public preview
257+
## Safety evaluation GPT-4, GPT-4o, GPT-4o-mini fine-tuning - public preview
252258

253259
[!INCLUDE [Safety evaluation](../includes/safety-evaluation.md)]
254260

‎articles/ai-services/openai/includes/fine-tuning-studio.md

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -30,13 +30,16 @@ The following models support fine-tuning:
3030
- `gpt-35-turbo` (1106)
3131
- `gpt-35-turbo` (0125)
3232
- `gpt-4` (0613)**<sup>*</sup>**
33+
- `gpt-4o` (2024-08-06)**<sup>*</sup>**
3334
- `gpt-4o-mini` (2024-07-18)**<sup>*</sup>**
3435

3536
**<sup>*</sup>** Fine-tuning for this model is currently in public preview.
3637

38+
Or you can fine tune a previously fine-tuned model, formatted as base-model.ft-{jobid}.
39+
40+
3741
Consult the [models page](../concepts/models.md#fine-tuning-models) to check which regions currently support fine-tuning.
3842

39-
If you plan to use `gpt-4` for fine-tuning, please refer to the [GPT-4 public preview safety evaluation guidance](#safety-evaluation-gpt-4-fine-tuning---public-preview)
4043

4144
## Review the workflow for Azure OpenAI Studio
4245

@@ -322,7 +325,7 @@ Here are some of the tasks you can do on the **Models** pane:
322325
When each training epoch completes a checkpoint is generated. A checkpoint is a fully functional version of a model which can both be deployed and used as the target model for subsequent fine-tuning jobs. Checkpoints can be particularly useful, as they can provide a snapshot of your model prior to overfitting having occurred. When a fine-tuning job completes you will have the three most recent versions of the model available to deploy.
323326

324327

325-
## Safety evaluation GPT-4 fine-tuning - public preview
328+
## Safety evaluation GPT-4, GPT-4o, and GPT-4o-mini fine-tuning - public preview
326329

327330
[!INCLUDE [Safety evaluation](../includes/safety-evaluation.md)]
328331

0 commit comments

Comments
 (0)