> ## Documentation Index
> Fetch the complete documentation index at: https://internal.nolano.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Configuration

> Configure model architecture and initialization

## Model Configuration

### ModelConfig

Configuration for model architecture and initialization.

<ParamField path="architecture" type="str" required>
  Model architecture specification. Supports major dense and MoE Hugging Face architectures including Qwen, LLaMA, Gemma.

  <Note>
    TSFM and other architectures coming soon.
  </Note>
</ParamField>

<ParamField path="init_method" type="str" default="normal">
  Weight initialization strategy:

  * `"none"`: Load from pre-trained model (Qwen/LLaMA/Gemma)
  * `"normal"`: Normal distribution initialization
  * `"xavier_uniform"`: Xavier uniform initialization
  * `"wang_init"`: Wang initialization method
</ParamField>

<ParamField path="model_path" type="str or None" default="None">
  Path to pre-trained model for continual training. Must be `None` if `init_method` is not `"none"`.
</ParamField>

<ParamField path="load_optimizer" type="bool or None" default="None">
  Whether to load optimizer state from checkpoint. Set to `True` for continual training from checkpoint.
</ParamField>

<ParamField path="precision" type="str" default="fp16">
  Model precision configuration:

  * `"binary"`: Binary precision (1-bit)
  * `"ternary"`: Ternary precision (1.58-bit)
  * `"int2"`: 2-bit integer precision
  * `"fp8"`: 8-bit floating point
  * `"mxfp4"`: 4-bit microscaling floating point
  * `"mxfp6"`: 6-bit microscaling floating point
  * `"ue8m0"`: 8-bit unsigned integer with 0 exponent bits
  * `"fp16"`: 16-bit floating point (default)
  * `"fp32"`: 32-bit floating point
</ParamField>

## Example Configurations

<CodeGroup>
  ```python Pre-trained Model theme={null}
  from pynolano import ModelConfig

  config = ModelConfig(
      architecture="Qwen/Qwen3-4B",
      init_method="none",
      model_path="./pretrained_model",
      load_optimizer=False,
      precision="fp16"
  )
  ```

  ```python From Scratch theme={null}
  from pynolano import ModelConfig

  config = ModelConfig(
      architecture="Qwen/Qwen3-4B",
      init_method="normal",
      precision="fp16"
  )
  ```

  ```python Continual Training theme={null}
  from pynolano import ModelConfig

  config = ModelConfig(
      architecture="Qwen/Qwen3-4B",
      init_method="none",
      model_path="./checkpoint/global_step_1000",
      load_optimizer=True,
      precision="fp16"
  )
  ```

  ```python Low Precision Training theme={null}
  from pynolano import ModelConfig

  config = ModelConfig(
      architecture="Qwen/Qwen3-4B",
      init_method="normal",
      precision="ternary"
  )
  ```
</CodeGroup>

## Supported Architectures

<Tabs>
  <Tab title="Dense Models">
    * **LLaMA Series**
    * **Qwen Series**
    * **Gemma Series**
  </Tab>

  <Tab title="Mixture of Experts (MoE)">
    * **Qwen MoE**
    * **DeepSeek**
    * **Mixtral**
  </Tab>

  <Tab title="Time Series Foundation Models">
    * **TSFM**: Time Series Foundation Models
    * **Chronos**: Amazon Chronos models
    * **TimesFM**: Google TimesFM
  </Tab>

  <Tab title="Low Precision Models">
    * **TriLM**: Ternary Language Models with 1.58-bit quantization
    * **BiLM**: Binary Language Models with 1-bit quantization
    * **Arbitrary-precision**: Language Models with Arbitrary Precision
  </Tab>
</Tabs>

you can request custom architecture if needed

<Warning>
  When using `init_method="none"`, you must provide a valid `model_path`. The model path should point to a compatible pre-trained model or checkpoint.
</Warning>