Instructions to use vivek078/AutodevvLM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use vivek078/AutodevvLM with Transformers:

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

pipe = pipeline("text-generation", model="vivek078/AutodevvLM", trust_remote_code=True)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("vivek078/AutodevvLM", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use vivek078/AutodevvLM with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf vivek078/AutodevvLM:F16
# Run inference directly in the terminal:
llama cli -hf vivek078/AutodevvLM:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf vivek078/AutodevvLM:F16
# Run inference directly in the terminal:
llama cli -hf vivek078/AutodevvLM:F16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf vivek078/AutodevvLM:F16
# Run inference directly in the terminal:
./llama-cli -hf vivek078/AutodevvLM:F16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf vivek078/AutodevvLM:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf vivek078/AutodevvLM:F16

Use Docker

docker model run hf.co/vivek078/AutodevvLM:F16

LM Studio
Jan

vLLM

How to use vivek078/AutodevvLM with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "vivek078/AutodevvLM"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vivek078/AutodevvLM",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/vivek078/AutodevvLM:F16

SGLang

How to use vivek078/AutodevvLM 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 "vivek078/AutodevvLM" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vivek078/AutodevvLM",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "vivek078/AutodevvLM" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "vivek078/AutodevvLM",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use vivek078/AutodevvLM with Ollama:
```
ollama run hf.co/vivek078/AutodevvLM:F16
```
Unsloth Desktop
Docker Model Runner
How to use vivek078/AutodevvLM with Docker Model Runner:
```
docker model run hf.co/vivek078/AutodevvLM:F16
```

Lemonade

How to use vivek078/AutodevvLM with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull vivek078/AutodevvLM:F16

Run and chat with the model

lemonade run user.AutodevvLM-F16

List all available models

lemonade list

Atomic Chat

AutodevvLM v0.1

AutodevvLM v0.1 is a ~10.06M parameter decoder-only Transformer trained from scratch on CPU. It follows GPT-2 conventions (fused QKV attention, learned position embeddings, pre-LN blocks) and is small enough to run comfortably on a laptop.

It ships with both a PyTorch/Hugging Face checkpoint (model.safetensors, load with trust_remote_code=True) and GGUF builds for llama.cpp.

Model details

Property	Value
Parameters	10,064,896
Architecture	Decoder-only Transformer (GPT-2 style)
Layers	10
Model dim (`n_embd`)	256
Attention heads	8
FFN inner dim	1024
Vocab size	4096
Context (`block_size`)	256
Params dtype	float32 (safetensors)

Files

File	Description
`config.json`	Model config (+ `auto_map` for `trust_remote_code`)
`configuration_autodevlm.py`	`AutoDevLMTinyGPTConfig` (remote code)
`modeling_autodevlm.py`	`AutoDevLMTinyGPTForCausalLM` (remote code)
`model.safetensors`	Weights (float32, ~38 MB)
`tokenizer.json`	Byte-level BPE tokenizer
`tokenizer_config.json`	Tokenizer special tokens
`autodevv-f16.gguf`	GGUF, F16 (~19.3 MB)
`autodevv-q4_k_m.gguf`	GGUF, Q4_K_M (~6.5 MB)

Usage (PyTorch / transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("<repo-id>", trust_remote_code=True)
tok = AutoTokenizer.from_pretrained("<repo-id>")

ids = tok("the tiny transformer", return_tensors="pt").input_ids
out = model.generate_text(ids, max_new_tokens=30, temperature=0.8, seed=0)
print(tok.decode(out[0]))

Usage (llama.cpp)

llama-cli -m autodevv-q4_k_m.gguf -p "the tiny transformer" -n 30 --temp 0.8 -s 0

Training

Trained from scratch on CPU (src/train.py, AdamW).
50 steps: train loss 1.74 → 0.05, val loss 2.37 → 1.24.
Byte-level tokenizer (ByteLevel(use_regex=False)), vocab padded to 4096.

Intended use & limitations

This is a tiny educational model trained on a very small corpus. Outputs are simple and domain-limited; it is not a general-purpose assistant and should be used for learning and experimentation only.

License

MIT.

Downloads last month: -

Safetensors

Model size

10.1M params

Tensor type

F32