# AI PLATFORM

#### Consideraciones

| Tema | Consideración |
|-|-|
| Máquinas |Máquinas disponibles soportan modelos hasta 2GB , los módelos son n1-standard-X (x Número de cpu hasta 32).Lo modelo con gpu solo aceptan TensorFlow|
| Ubicación Máquinas| Las máquinas están en us-central1|
| Quota| 100 modelos por proyecto|
|||

##### Cantidad aceptada de gpus por máquina.

##### Compute Engine n1

| |Compute Engine N1 |
|-|-|
| Regions| us-central1|
| Types of ML artifacts| TensorFlow SavedModels |
| Runtime versions| 1.1 or later|
| Max Model size| 2GB |

**Deployment URI**

You must provide the URI of the Cloud Storage location where your SavedModel is stored. AI Platform pulls the model from this location and deploys it. This parameter is called --origin in the gcloud ai-platform versions create command. If you deploy a custom prediction routine (beta), you may provide the URI of a Cloud Storage directory containing any artifacts your model version uses for prediction, not just a SavedModel.

gcloud ai-platform versions create

#### Consultas a la API

No es posible darles a los datos binarios el formato de strings codificadas en UTF-8 que admite JSON. Si tienes datos binarios en tus entradas, debes usar la codificación base64 para representarlos.

En Python 3.5, la codificación base64 genera una secuencia de bytes. Debes convertir esto en una string para que pueda serializarse con JSON.

{'image_bytes': {'b64': base64.b64encode(jpeg_data).decode()}}

En tu código de modelo TensorFlow, debes nombrar los alias de los tensores de entrada y salida binarios de modo que finalicen con “_bytes

#### Solicitus predicciones

Envía tus instancias de datos de entrada como una string JSON en una solicitud de predicción para solicitar una predicción en línea. Si deseas dar formato al cuerpo de la solicitud y la respuesta, consulta los detalles de la solicitud de predicción.

Si no especificas una versión del modelo, se usa la versión predeterminada del modelo en tu solicitud de predicción.


url: https://cloud.google.com/ml-engine/docs/online-predict#requesting_logs_for_online_prediction_requests

### Importante

**ML framework support**

If you use one of the Compute Engine (N1) machine types, you must create your model version with a TensorFlow SavedModel and specify TENSORFLOW for the framework field.

**Automatic scaling**

If you use a Compute Engine (N1) machine type with automatic scaling, your model version must always have at least one node running. In other words, the version's autoScaling.minNodes field defaults to 1 and cannot be less than 1.

If you use GPUs for your model version, you cannot use automatic scaling. You must use manual scaling.

**Manual scaling**

If you use a Compute Engine (N1) machine type with manual scaling, you can update the number of prediction nodes running at any time by using the projects.models.versions.patch API method.

**Runtime version support**

If you use a Compute Engine (N1) machine types, you must use runtime version 1.11 or later for your model version.

**Runtime version list**

AI Platform Training uses images, based on runtime version designation, to configure cloud resources to service your training and prediction requests.

|version |support for training |
|-|-|
|TensorFlow 1.13 |Starting on March 6, 2020, you can no longer create training jobs, batch prediction jobs, or model versions that use this runtime version. On March 6, 2021, AI Platform Prediction deletes all your model versions that use this runtime version. |
|TensorFlow 1.14.0 | Starting on July 19, 2020, you can no longer create training jobs, batch prediction jobs, or model versions that use this runtime version.On July 19, 2021, AI Platform Prediction deletes all your model versions that use this runtime version.|
|TensorFlow 1.15 | Starting on December 19, 2020, you can no longer create training jobs, batch prediction jobs, or model versions that use this runtime version. On December 19, 2021, AI Platform Prediction deletes all your model versions that use this runtime version. |

https://cloud.google.com/blog/products/ai-machine-learning/now-you-can-train-ml-models-faster-and-lower-cost-cloud-tpu-pods?utm_source=cloudaiblog&utm_medium=blog&utm_campaign=tpuv2v3podsbeta