Instructions to use ConeML/coneml-348m-gamma with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ConeML/coneml-348m-gamma with Transformers:

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

pipe = pipeline("text-generation", model="ConeML/coneml-348m-gamma")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ConeML/coneml-348m-gamma")
model = AutoModelForCausalLM.from_pretrained("ConeML/coneml-348m-gamma")
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 Settings

vLLM

How to use ConeML/coneml-348m-gamma with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ConeML/coneml-348m-gamma"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ConeML/coneml-348m-gamma",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ConeML/coneml-348m-gamma

SGLang

How to use ConeML/coneml-348m-gamma 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 "ConeML/coneml-348m-gamma" \
    --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": "ConeML/coneml-348m-gamma",
		"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 "ConeML/coneml-348m-gamma" \
        --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": "ConeML/coneml-348m-gamma",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ConeML/coneml-348m-gamma with Docker Model Runner:
```
docker model run hf.co/ConeML/coneml-348m-gamma
```

RandomMountainMan commited on 6 days ago

Commit

ae48ca7

verified ·

1 Parent(s): 556961e

Scrub training-lineage detail from card (step count / checkpoint / SFT stage)

Browse files

Files changed (1) hide show

README.md +3 -4

README.md CHANGED Viewed

@@ -36,10 +36,9 @@ excluding token embeddings. Qwen2.5-0.5B is 494M total and roughly 360M non-embe
 Gamma therefore runs at ~70% of the total parameters and ~88% of the non-embedding core of the model it is measured
 against below.
-Training budget disclosure: Gamma was selected from a pretraining checkpoint at 210,000 steps × 32,768 tokens per
-step ≈ 6.9B pretraining tokens, followed by a small SFT stage. Qwen2.5 reports an 18-trillion-token pretraining
-corpus at the series level; per-size splits are not published, but on any reading the results below are achieved
-with roughly three orders of magnitude less training data.
 ## Results at a Glance

 Gamma therefore runs at ~70% of the total parameters and ~88% of the non-embedding core of the model it is measured
 against below.
+Training budget disclosure: Gamma's disclosed pretraining budget is approximately 6.9B tokens. Qwen2.5 reports an
+18-trillion-token pretraining corpus at the series level; per-size splits are not published, but on any reading the
+results below are achieved with roughly three orders of magnitude less training data.
 ## Results at a Glance