Upload folder using huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- code-generation
|
| 5 |
+
- plc
|
| 6 |
+
- iec-61131-3
|
| 7 |
+
- structured-text
|
| 8 |
+
- gemma
|
| 9 |
+
- phi
|
| 10 |
+
- industrial-automation
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# PLC Code Generation Models (IEC 61131-3)
|
| 14 |
+
|
| 15 |
+
This repository contains fine-tuned checkpoints for generating Programmable Logic Controller (PLC) code in **Structured Text (IEC 61131-3)** format. These models have been explicitly trained and benchmarked for air-gapped, edge-hardware deployments in industrial settings.
|
| 16 |
+
|
| 17 |
+
## Models Included
|
| 18 |
+
|
| 19 |
+
This repository hosts weights for two distinct model families:
|
| 20 |
+
|
| 21 |
+
### 1. Gemma 3 (4B) - `Gemma3-PLC-4B/`
|
| 22 |
+
A fine-tuned version of Google's Gemma 3 (4B parameter) model. It was fine-tuned for high-accuracy complex logic generation and benchmarked for latency and throughput on edge GPUs (e.g., NVIDIA A4000).
|
| 23 |
+
- **Folders Included**: `model_merged`, `model_merged_hacked`, and intermediate `checkpoints_*`.
|
| 24 |
+
|
| 25 |
+
### 2. Phi - `Phi-PLC-Lora/`
|
| 26 |
+
A highly efficient fine-tune of Microsoft's Phi model. Provided as both merged weights and standalone LoRA adapters.
|
| 27 |
+
- **Folders Included**: `model_phi_merged`, `model_phi_lora`.
|
| 28 |
+
|
| 29 |
+
## Intended Use
|
| 30 |
+
|
| 31 |
+
- **Task**: Natural Language to PLC Structured Text (IEC 61131-3).
|
| 32 |
+
- **Environment**: Optimized for local, on-premise edge inference where internet access is restricted (air-gapped factory environments).
|
| 33 |
+
- **Users**: Control Systems Engineers, Automation Programmers, and Robotics Researchers.
|
| 34 |
+
|
| 35 |
+
## Training Data
|
| 36 |
+
The models were trained on a highly curated dataset consisting of:
|
| 37 |
+
- Standardized open-source PLC libraries (e.g., OSCAT).
|
| 38 |
+
- Custom verified industrial control logic datasets.
|
| 39 |
+
|
| 40 |
+
## Benchmarking & Performance
|
| 41 |
+
These models were evaluated against zero-shot baselines for both generation accuracy and inference speed. They achieve state-of-the-art performance for localized IEC 61131-3 generation, significantly outperforming base models on complex automation tasks. Detailed benchmarking scripts and data parsing logic can be found in the associated GitHub repository.
|