README.md

metadata

license: apache-2.0
base_model:
  - Qwen/Qwen2.5-Coder-1.5B-Instruct
tags:
  - gguf
  - ollama
  - code
  - text-generation
  - reactjs
  - nodejs
model_type: qwen2
pipeline_tag: text-generation
quantized_by: llama-quantize

Videxpulse-Coder-1.5B-GGUF

This is a fine-tuned version of Qwen 1.5B, quantized to Q4_K_M GGUF format. It is optimized specifically for full-stack web development.

Specialization

• ReactJS, NodeJS, ExpressJS
• MySQL, PostgreSQL
• WordPress, PHP

How to use with Ollama

Quick Start

1. Download the videxpulse-coder-1.5b-q4_k_m.gguf and the Modelfile.
2. Run the following command in your terminal to create the model:

   ollama create Videxpulse-Coder -f ./Modelfile
   

3. Run the model:

   ollama run Videxpulse-Coder
   

Creating a Modelfile

If you need to create or customize the Modelfile, follow these instructions:

1. Create a new file named Modelfile (no extension) in the directory containing your GGUF model file.

2. Add the base model reference:

   FROM ./videxpulse-coder-1.5b-q4_k_m.gguf
   

   Replace videxpulse-coder-1.5b-q4_k_m.gguf with your actual GGUF file name if different.

3. Configure model parameters (optional but recommended):

   PARAMETER temperature 0.2
   PARAMETER stop "<|im_start|>"
   PARAMETER stop "<|im_end|>"
   

   • temperature: Controls randomness in responses (0.0-1.0). Lower values (0.2) produce more deterministic code.
   • stop: Defines tokens where the model should stop generating.

4. Define the chat template (required for proper formatting):

   TEMPLATE """{{ if .System }}<|im_start|>system
   {{ .System }}<|im_end|>
   {{ end }}{{ if .Prompt }}<|im_start|>user
   {{ .Prompt }}<|im_end|>
   {{ end }}<|im_start|>assistant
   {{ .Response }}<|im_end|>"""
   

   This template ensures proper message formatting for the Qwen model architecture.

5. Set the system prompt (customize based on your needs):

   SYSTEM """You are Videxpulse-Coder, an advanced open-source AI assistant fine-tuned specifically to write elite ReactJS, NodeJS, expressJS, MYSQL, Postgresql, WordPress, PHP code. You are an independent AI assistant optimized for software engineering."""
   

6. Save the Modelfile and verify it's in the same directory as your GGUF file.

7. Create the Ollama model using the Modelfile:

   ollama create Videxpulse-Coder -f ./Modelfile
   

Complete Modelfile Example

Here's a complete example of a ready-to-use Modelfile:

FROM ./videxpulse-coder-1.5b-q4_k_m.gguf

PARAMETER temperature 0.2
PARAMETER stop "<|im_start|>"
PARAMETER stop "<|im_end|>"

TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>"""

SYSTEM """You are Videxpulse-Coder, an advanced open-source AI assistant fine-tuned specifically to write elite ReactJS, NodeJS, expressJS, MYSQL, Postgresql, WordPress, PHP code. You are an independent AI assistant optimized for software engineering."""

Hardware & Estimation Requirements

This model is heavily optimized for resource-constrained local environments. Below are the resource estimations for efficient execution:

📊 RAM & VRAM Footprint

• Model File Size: ~1.02 GB
• Minimum VRAM Required: ~1.40 GB (For full GPU offloading with 2k context window)
• Minimum System RAM: 4 GB (If running purely on CPU/System memory)

💻 Supported Hardware Classes

• Apple Silicon: M1, M2, M3, M4 series (Runs entirely in unified memory via Ollama)
• NVIDIA GPUs: GTX 10-series, RTX 20/30/40/50 series (Requires >2GB VRAM for full offload)
• CPU Architectures: x86_64 with AVX2 instruction set support