Instructions to use mehul945/frappe-coder-3b-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mehul945/frappe-coder-3b-lora with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="mehul945/frappe-coder-3b-lora")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("mehul945/frappe-coder-3b-lora", device_map="auto")

PEFT
How to use mehul945/frappe-coder-3b-lora with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use mehul945/frappe-coder-3b-lora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mehul945/frappe-coder-3b-lora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mehul945/frappe-coder-3b-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mehul945/frappe-coder-3b-lora

SGLang

How to use mehul945/frappe-coder-3b-lora with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "mehul945/frappe-coder-3b-lora" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mehul945/frappe-coder-3b-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "mehul945/frappe-coder-3b-lora" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mehul945/frappe-coder-3b-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mehul945/frappe-coder-3b-lora with Docker Model Runner:
```
docker model run hf.co/mehul945/frappe-coder-3b-lora
```

frappe-coder-3b-lora

A LoRA fine-tuned version of Qwen/Qwen2.5-Coder-3B-Instruct, specialized for Frappe Framework and ERPNext code generation.

Model Details

Model Description

Developed by: Mehul Talpada
Model type: Causal LM with LoRA adapter (PEFT)
Language: English
License: Apache 2.0
Base model: Qwen/Qwen2.5-Coder-3B-Instruct
Fine-tuning method: LoRA (Low-Rank Adaptation)
Domain: Frappe Framework / ERPNext application development

This model is fine-tuned to assist developers working with the Frappe Framework and ERPNext ERP platform — generating DocTypes, controllers, server scripts, REST API handlers, hooks, fixtures, and other Frappe-specific Python/JS code.

Uses

Direct Use

Generating Frappe DocType definitions and controller logic
Writing ERPNext server scripts and scheduled tasks
Auto-completing Frappe Python APIs (frappe.get_doc, frappe.db.*, etc.)
Generating REST API endpoints and whitelisted methods
Writing Frappe hooks, fixtures, and patch files

Out-of-Scope Use

General-purpose coding outside the Frappe/ERPNext ecosystem (may still work, but not optimized)
Production use without human review of generated code
Security-sensitive code without manual audit

How to Get Started

Load the adapter (requires base model)

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

BASE_MODEL = "Qwen/Qwen2.5-Coder-3B-Instruct"
ADAPTER    = "your-username/frappe-coder-3b-lora"

tokenizer = AutoTokenizer.from_pretrained(ADAPTER)
model     = AutoModelForCausalLM.from_pretrained(BASE_MODEL, torch_dtype="auto")
model     = PeftModel.from_pretrained(model, ADAPTER)
model.eval()

Run inference

prompt = """### Instruction:
Write a Frappe controller method that sends an email notification when a Sales Order is submitted.

### Response:
"""

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512, temperature=0.2)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Training Details

Training Data

Fine-tuned on a custom dataset (frappe_dataset) built from:

Frappe Framework source repositories
ERPNext application source code
Instruction-response pairs covering common Frappe development tasks

Training Procedure

Parameter	Value
Base model	Qwen2.5-Coder-3B-Instruct
Fine-tune method	LoRA (PEFT)
Precision	bf16 mixed precision
Framework	Hugging Face Transformers + PEFT
Hardware	Kaggle GPU (T4/P100)

LoRA Configuration

Parameter	Value
`r` (rank)	16
`lora_alpha`	32
`target_modules`	q_proj, v_proj
`lora_dropout`	0.05
`bias`	none

Update these values with your actual training config.

Limitations

The model may hallucinate Frappe API methods that don't exist — always verify against Frappe docs
Trained on a limited dataset; rare or advanced Frappe patterns may not be handled well
Not tested on Frappe v15+ specific APIs (verify compatibility with your version)

Citation

If you use this model, please cite the base model:

@misc{qwen2.5coder,
  title  = {Qwen2.5-Coder},
  author = {Qwen Team, Alibaba Cloud},
  year   = {2024},
  url    = {https://huggingface.co/Qwen/Qwen2.5-Coder-3B-Instruct}
}


---

**To use this:** paste it into the `README.md` at the root of your HF repo. Fill in your actual LoRA hyperparameters (rank, alpha, target modules) from your training config — those are the most important fields to get right.

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for mehul945/frappe-coder-3b-lora

Base model

Qwen/Qwen2.5-3B

Finetuned

Qwen/Qwen2.5-Coder-3B

Finetuned

Qwen/Qwen2.5-Coder-3B-Instruct

Adapter

(78)

this model