> ## Documentation Index
> Fetch the complete documentation index at: https://forge-64364c0e.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Training

> Fine-tune models with LoRA, QLoRA, SFT, DPO, or Full training — plus pure-Rust layer surgery

# Training (06)

The Training module provides two modes: **Fine-Tune** for GPU-accelerated model training, and **Layer Surgery** for pure-Rust tensor operations that don't require Python or a GPU.

<img src="https://mintcdn.com/forge-64364c0e/ndaDrHy7hzGW88bh/images/light/traning-finetuning-light.png?fit=max&auto=format&n=ndaDrHy7hzGW88bh&q=85&s=d66e4e7d34325da57dd87697229bf2b4" alt="Training Fine-tune" width="3440" height="1406" data-path="images/light/traning-finetuning-light.png" />

<Note>
  Fine-tuning requires **Python 3.10+** and a one-time environment setup. Layer surgery works immediately with no dependencies.
</Note>

## Fine-Tune Mode

### Training Methods

| Method    | Description                                                        | VRAM Needed |
| --------- | ------------------------------------------------------------------ | :---------: |
| **LoRA**  | Low-Rank Adaptation — efficient adapter training                   |   6–12 GB   |
| **QLoRA** | Quantized LoRA — 4-bit base model with LoRA adapters               |    4–8 GB   |
| **SFT**   | Supervised Fine-Tuning — standard training on instruction datasets |   8–24 GB   |
| **DPO**   | Direct Preference Optimization — learn from chosen/rejected pairs  |   8–24 GB   |
| **Full**  | Full parameter update — maximum quality, highest VRAM              |   16–48 GB  |

### VRAM Presets

<CardGroup cols={4}>
  <Card title="LOW VRAM" icon="battery-quarter">
    **\~4 GB** — QLoRA, rank 8, 256 seq length
  </Card>

  <Card title="BALANCED" icon="scale-balanced">
    **\~6 GB** — QLoRA, rank 16, 512 seq length
  </Card>

  <Card title="QUALITY" icon="star">
    **\~12 GB** — LoRA, rank 32, 1024 seq length
  </Card>

  <Card title="MAX QUALITY" icon="crown">
    **\~24 GB** — LoRA, rank 64, 2048 seq length
  </Card>
</CardGroup>

### Hyperparameters

Full control over all training parameters:

| Parameter             | Description                               |
| --------------------- | ----------------------------------------- |
| Learning Rate         | Step size for gradient descent            |
| Epochs                | Number of full passes through the dataset |
| Batch Size            | Samples per training step                 |
| Gradient Accumulation | Effective batch size multiplier           |
| Max Sequence Length   | Maximum token length per sample           |
| Warmup Steps          | Linear warmup steps at start              |
| Weight Decay          | L2 regularization strength                |
| Save Steps            | Checkpoint save interval                  |
| LoRA Rank             | Rank of low-rank matrices (8–64)          |
| LoRA Alpha            | Scaling factor for LoRA                   |
| LoRA Dropout          | Dropout probability on adapters           |
| Quantization Bits     | 4 or 8 (for QLoRA)                        |
| DPO Beta              | KL penalty coefficient (for DPO)          |

### Capability-Targeted Layer Selection

Instead of fine-tuning all layers, target specific model capabilities:

| Capability                | Affected Layers | What It Trains                 |
| ------------------------- | :-------------: | ------------------------------ |
| **Tool Calling**          |    Upper-mid    | API/function calling ability   |
| **Reasoning / CoT**       |    Mid-upper    | Chain-of-thought reasoning     |
| **Code Generation**       |    Upper-mid    | Code writing and understanding |
| **Mathematics**           |       Mid       | Mathematical reasoning         |
| **Multilingual**          |    Early-mid    | Multi-language support         |
| **Instruction Following** |       Mid       | Adherence to instructions      |
| **Safety & Alignment**    |      Final      | Safety behavior adjustment     |

### Target Module Detection

ForgeAI auto-detects available LoRA target modules from the model architecture:

| Module      | Component                     |
| ----------- | ----------------------------- |
| `q_proj`    | Query projection (Attention)  |
| `k_proj`    | Key projection (Attention)    |
| `v_proj`    | Value projection (Attention)  |
| `o_proj`    | Output projection (Attention) |
| `gate_proj` | Gate projection (MLP)         |
| `up_proj`   | Up projection (MLP)           |
| `down_proj` | Down projection (MLP)         |

### Dataset Support

* **Auto-detection** of templates: Alpaca, ShareGPT, ChatML, DPO pairs, Text, Prompt/Completion
* **Formats**: JSON, JSONL, CSV, Parquet
* **Preview**: View dataset rows and column structure before training

### Live Training Dashboard

During training, a real-time dashboard shows:

* Epoch and step progress
* Loss value with step-by-step loss history
* Learning rate (with warmup visualization)
* GPU VRAM usage
* ETA / time remaining
* Option to merge adapter back into base model after completion

### Workflow

<Steps>
  <Step title="Setup environment">
    First time only: ForgeAI checks for Python, creates a venv, installs PyTorch + PEFT + TRL.
  </Step>

  <Step title="Select model">
    Browse for a GGUF or SafeTensors model (or folder).
  </Step>

  <Step title="Select dataset">
    Browse for a training dataset. ForgeAI auto-detects format and template.
  </Step>

  <Step title="Choose method and preset">
    Pick a training method and VRAM preset. Optionally target specific capabilities.
  </Step>

  <Step title="Train">
    Click **START TRAINING** and monitor real-time progress.
  </Step>
</Steps>

***

## Layer Surgery Mode

<img src="https://mintcdn.com/forge-64364c0e/ndaDrHy7hzGW88bh/images/light/traning-surgery.png?fit=max&auto=format&n=ndaDrHy7hzGW88bh&q=85&s=98aa5289390ac5c24013d8523bdf3d41" alt="Layer Surgery" width="3440" height="1406" data-path="images/light/traning-surgery.png" />

Pure Rust tensor operations — **no Python or GPU required**.

### Operations

| Operation            | Description                                          |
| -------------------- | ---------------------------------------------------- |
| **Remove Layers**    | Select and strip layers to reduce model size         |
| **Duplicate Layers** | Clone layers at specific positions to increase depth |

### Features

* **Rich Layer Table** — memory breakdown per layer with component bars (attention/MLP/norm %)
* **Tensor-Level Inspection** — expand any layer to see every tensor's dtype, shape, and memory
* **Surgery Preview** — shows final layer count before execution
* **Format Support** — works with both SafeTensors directories and GGUF files
* **Auto-Update** — automatically updates `config.json` / GGUF metadata with new layer counts

### Workflow

<Steps>
  <Step title="Select model">
    Browse for a GGUF file or SafeTensors folder
  </Step>

  <Step title="Load layer details">
    Click **LOAD LAYER DETAILS** to see all layers with tensor-level breakdown
  </Step>

  <Step title="Select operations">
    Check layers to **remove**, or add layers to **duplicate** at specific positions
  </Step>

  <Step title="Execute">
    Click **RUN SURGERY**. The output is a new model file — the original is never modified.
  </Step>
</Steps>