Instructions to use ganjakha/codium-bd-coder-v3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ganjakha/codium-bd-coder-v3 with Transformers:

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

pipe = pipeline("text-generation", model="ganjakha/codium-bd-coder-v3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ganjakha/codium-bd-coder-v3")
model = AutoModelForCausalLM.from_pretrained("ganjakha/codium-bd-coder-v3", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ganjakha/codium-bd-coder-v3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ganjakha/codium-bd-coder-v3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ganjakha/codium-bd-coder-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ganjakha/codium-bd-coder-v3

SGLang

How to use ganjakha/codium-bd-coder-v3 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 "ganjakha/codium-bd-coder-v3" \
    --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": "ganjakha/codium-bd-coder-v3",
		"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 "ganjakha/codium-bd-coder-v3" \
        --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": "ganjakha/codium-bd-coder-v3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ganjakha/codium-bd-coder-v3 with Docker Model Runner:
```
docker model run hf.co/ganjakha/codium-bd-coder-v3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Codium-BD-Coderv3

Coding-focused Bangla + English model by Niloy Chandra Deb. Full merged weights (usable directly, or as a base for further fine-tuning).

Lineage: Qwen3.5-4B (please follow the Qwen license and keep attribution) + LoRA merge + identity tuning + v3.1 fine-tune
v3.1 fine-tune data: Bangla instruction sets (HydraIndicLM/bengali_alpaca_dolly_67k, iamshnoo/alpaca-cleaned-bengali, BanglaLLM/bangla-alpaca-orca, OdiaGenAI/all_combined_bengali_252k), real-world HTML/CSS from ronantakizawa/webui (MIT label; underlying code has mixed open-source licenses), Tailwind pages from HuggingFaceM4/WebSight v0.2 (CC-BY-4.0, https://arxiv.org/abs/2403.09029)
Previous version: branch v3.0-backup
Needs transformers>=4.57

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("ganjakha/codium-bd-coder-v3")
model = AutoModelForCausalLM.from_pretrained("ganjakha/codium-bd-coder-v3", dtype="auto", device_map="auto")
msgs = [{"role": "user", "content": "Make a responsive landing page for a coffee shop (single HTML file)."}]
inp = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt", return_dict=True).to(model.device)
print(tok.decode(model.generate(**inp, max_new_tokens=800)[0][inp["input_ids"].shape[1]:], skip_special_tokens=True))

Limitations

Review generated code before use. No formal benchmarks have been run.

Downloads last month: -

Safetensors

Model size

4B params

Tensor type

F16

Model tree for ganjakha/codium-bd-coder-v3

Unable to build the model tree, the base model loops to the model itself. Learn more.

Datasets used to train ganjakha/codium-bd-coder-v3

Paper for ganjakha/codium-bd-coder-v3

Unlocking the conversion of Web Screenshots into HTML Code with the WebSight Dataset

Paper • 2403.09029 • Published Mar 14, 2024 • 57