import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';

# Quick Start Guide 

### 1. Create a Model Repository

-  Navigate to **Inference → Model Repository** in the TIR Dashboard
- Click **CREATE REPOSITORY**
- Enter a **repository name** (e.g., `my-llm-model`)
- Select **model type**:
   - **PyTorch** - For PyTorch models (`.pth`, `.pt`, Hugging Face models)
   - **Triton** - For Triton Inference Server model repositories
   - **Custom** - For other model formats
- Choose **storage type**:
   - **New EOS Bucket** - Automatically provisions a new bucket (recommended)
   - **Existing EOS Bucket** - Use an existing bucket
   - **External EOS Bucket** - Connect to an external bucket with credentials
- Click **Create**

<!-- ![Create Model Repository](images/create_model_repo.png) -->

### 2: Configure Access

After creation, you'll receive:
- **Bucket name**
- **Access key**
- **Secret key**
- **Endpoint URL** (typically `https://objectstore.e2enetworks.net`)

:::tip
Model Repositories are backed by E2E Object Storage (EOS). If you have not used EOS storage before, please read [Object Storage](/docs/myaccount/storage/object_storage/) first.
:::

### 3: Upload Model Files

#### Method 1: Using Minio CLI (Recommended)

Install Minio CLI:

```bash
# macOS
brew install minio/stable/mc

# Linux
wget https://dl.min.io/client/mc/release/linux-amd64/mc
chmod +x mc
sudo mv mc /usr/local/bin/
```

Configure the client:

```bash
mc config host add <alias-name> https://objectstore.e2enetworks.net <access-key> <secret-key>
```

Upload model files:

```bash
# Upload entire directory
mc cp -r /path/to/model/* <alias-name>/<bucket-name>/

# Upload specific version
mc cp -r /path/to/model/* <alias-name>/<bucket-name>/v1/

# Upload Hugging Face model snapshot
mc cp -r ~/.cache/huggingface/hub/<model-name>/* <alias-name>/<bucket-name>/
```  

#### Method 2: Using Python SDK

Step 1: Install the SDK
```python
pip install e2enetworks
```
Step 2: Initialize the SDK
```python
from e2enetworks.cloud import tir
from e2enetworks.constants import DELHI_LOCATION, CHENNAI_LOCATION

# Choose the required location
location = DELHI_LOCATION

# Initialize TIR SDK
# Get API Key and Access Token from Projects → API Tokens
tir.init(
    api_key="<API_KEY>",
    access_token="<ACCESS_TOKEN>",
    location=location
)

```
Step 3: Create a Model Repository client
```python
model_repo_client = tir.Models(
    project=<PROJECT_ID>
)
```
Step 4: Upload a model
```python
# Upload all files from a local directory to the model repository
model_repo_client.push_model(
    model_path="./model-dir",
    prefix="v1",
    model_id=<MODEL_ID>
)
```

#### Method 3: Using TIR Notebooks

In a TIR Notebook, use the SDK or CLI directly:

Step 1: Install the SDK
```python
pip install e2enetworks
```
Step 2: Run commands in a TIR Notebook
```python
from e2enetworks.cloud import tir

# Set location
location = "Delhi"

# Create model repository client
model_repo_client = tir.Models(location=location)

# Upload a model version from a local directory
model_repo_client.push_model(
    model_path="<LOCAL_MODEL_DIRECTORY>",
    prefix="<VERSION_PREFIX>",
    model_id=2297
)

# Download model files from the repository
model_repo_client.download_model(
    model_id=2297,
    local_path="<LOCAL_DOWNLOAD_PATH>",
    prefix="<VERSION_PREFIX>"
)

```

### 4: Verify Upload

1. Navigate to your Model Repository in the dashboard
2. Click on the **Model Files** tab
3. Verify your files are uploaded correctly


### 5: Use with Model Endpoints

1. Create a **Model Endpoint** (Inference → Model Endpoints)
2. Select **Link with Model Repository**
3. Choose your repository from the list
4. Configure the model path (e.g., `/` for root, `/v1` for versioned models)
5. Launch the endpoint

The model files will be automatically downloaded to the endpoint container during startup.

---


## Supported Frameworks

Model Repositories support storing models from various ML frameworks:

| Framework | File Formats | Use Case |
|-----------|-------------|----------|
| **PyTorch** | `.pth`, `.pt`, `.pkl` | PyTorch checkpoints, Hugging Face models, custom PyTorch models |
| **TensorFlow** | SavedModel, `.h5`, `.pb` | TensorFlow SavedModels, Keras models |
| **Triton Inference Server** | Triton model repository structure | Triton models with configuration files |
| **Custom Formats** | Any file structure | ONNX models, TensorRT engines, or any custom format |

### Example Directory Structures

<Tabs>
  <TabItem value="pytorch" label="PyTorch" default>

```
my-pytorch-model/
├── config.json
├── pytorch_model.bin
├── tokenizer.json
└── tokenizer_config.json
```

  </TabItem>
  <TabItem value="tensorflow" label="TensorFlow">

```
my-tensorflow-model/
├── saved_model.pb
├── variables/
│   ├── variables.data-00000-of-00001
│   └── variables.index
└── assets/
```

  </TabItem>
  <TabItem value="triton" label="Triton">

```
my-triton-model/
├── config.pbtxt
└── 1/
    └── model.graphdef
```

  </TabItem>
  <TabItem value="custom" label="Custom">

```
my-custom-model/
├── model.onnx
├── config.yaml
└── assets/
```

  </TabItem>
</Tabs>

---


---