Instructions to use moelanoby/phi-3-M3-coder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use moelanoby/phi-3-M3-coder with Transformers:

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

pipe = pipeline("text-generation", model="moelanoby/phi-3-M3-coder", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("moelanoby/phi-3-M3-coder", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("moelanoby/phi-3-M3-coder", trust_remote_code=True)
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use moelanoby/phi-3-M3-coder with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "moelanoby/phi-3-M3-coder"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "moelanoby/phi-3-M3-coder",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/moelanoby/phi-3-M3-coder

SGLang

How to use moelanoby/phi-3-M3-coder 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 "moelanoby/phi-3-M3-coder" \
    --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": "moelanoby/phi-3-M3-coder",
		"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 "moelanoby/phi-3-M3-coder" \
        --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": "moelanoby/phi-3-M3-coder",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use moelanoby/phi-3-M3-coder with Docker Model Runner:
```
docker model run hf.co/moelanoby/phi-3-M3-coder
```

🚩 Report: Spam

by kodecreer - opened Jul 5, 2025

Discussion

kodecreer

Jul 5, 2025

It's just Phi-3 with fake claims.

Doctor-Chad-PhD

Jul 7, 2025

I'm just confused what this is. It says in the upload notes:

Upload fine-tuned model with vector memory layer

So ok, let's say that it's fine tuned, then why does the shasum of the second safetensors file match the original phi3 file?

https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/model-00002-of-00002.safetensors

SHA256: 3f311787aa136e858556caa8543015161edcad85ba81b6a36072443d7fa73c87

https://huggingface.co/moelanoby/phi-3-M3-coder/blob/main/model-00002-of-00002.safetensors

SHA256: 3f311787aa136e858556caa8543015161edcad85ba81b6a36072443d7fa73c87

I will say that the first safetensors appears as unique,... but is it even possible to finetune just a single safetensors file?
That would be like finetuning just half the layers, I suppose.

To be fair, I think it's cool that the author is trying new things, so I do respect that part.

But it would be good if the description was more clear as to what it is.
The scores are misleading too, based on the results that I've seen by other people.

win10

Jul 8, 2025

I'm just confused what this is. It says in the upload notes:

Upload fine-tuned model with vector memory layer

So ok, let's say that it's fine tuned, then why does the shasum of the second safetensors file match the original phi3 file?

https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/blob/main/model-00002-of-00002.safetensors
SHA256: 3f311787aa136e858556caa8543015161edcad85ba81b6a36072443d7fa73c87
https://huggingface.co/moelanoby/phi-3-M3-coder/blob/main/model-00002-of-00002.safetensors
SHA256: 3f311787aa136e858556caa8543015161edcad85ba81b6a36072443d7fa73c87
I will say that the first safetensors appears as unique,... but is it even possible to finetune just a single safetensors file?
That would be like finetuning just half the layers, I suppose.

To be fair, I think it's cool that the author is trying new things, so I do respect that part.

But it would be good if the description was more clear as to what it is.
The scores are misleading too, based on the results that I've seen by other people.

It’s actually very simple. The author only fine-tuned the memory layer to adapt the model to the layer

win10

Jul 8, 2025

Many people don’t even take the time to look at the author’s architecture code.Just abuse false claims, it's ridiculous

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment