Instructions to use stage-babylm/llama-64-8L with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use stage-babylm/llama-64-8L with Transformers:

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

pipe = pipeline("text-generation", model="stage-babylm/llama-64-8L")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("stage-babylm/llama-64-8L")
model = AutoModelForCausalLM.from_pretrained("stage-babylm/llama-64-8L", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use stage-babylm/llama-64-8L with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "stage-babylm/llama-64-8L"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "stage-babylm/llama-64-8L",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/stage-babylm/llama-64-8L

SGLang

How to use stage-babylm/llama-64-8L 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 "stage-babylm/llama-64-8L" \
    --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": "stage-babylm/llama-64-8L",
		"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 "stage-babylm/llama-64-8L" \
        --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": "stage-babylm/llama-64-8L",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use stage-babylm/llama-64-8L with Docker Model Runner:
```
docker model run hf.co/stage-babylm/llama-64-8L
```

llama-64-8L

This model is a fine-tuned version of on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 2.0465

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0014
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.95) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: warmup_stable_decay
lr_scheduler_warmup_steps: 0.05
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0	0	6.9217
4.3489	0.0124	500	4.0773
3.0299	0.0248	1000	2.9862
2.7171	0.0372	1500	2.6991
2.5988	0.0497	2000	2.5863
2.5024	0.0621	2500	2.4978
2.4435	0.0745	3000	2.4402
2.4084	0.0869	3500	2.4038
2.3855	0.0993	4000	2.3786
2.3679	0.1117	4500	2.3482
2.3520	0.1241	5000	2.3395
2.3365	0.1366	5500	2.3252
2.3176	0.1490	6000	2.3101
2.2935	0.1614	6500	2.3008
2.2976	0.1738	7000	2.2926
2.2780	0.1862	7500	2.2898
2.2805	0.1986	8000	2.2810
2.2808	0.2110	8500	2.2824
2.2628	0.2234	9000	2.2688
2.2647	0.2359	9500	2.2616
2.2659	0.2483	10000	2.2593
2.2539	0.2607	10500	2.2518
2.2536	0.2731	11000	2.2498
2.2387	0.2855	11500	2.2489
2.2478	0.2979	12000	2.2455
2.2398	0.3103	12500	2.2473
2.2349	0.3228	13000	2.2372
2.2437	0.3352	13500	2.2355
2.2305	0.3476	14000	2.2317
2.2295	0.3600	14500	2.2314
2.2185	0.3724	15000	2.2327
2.2292	0.3848	15500	2.2258
2.2232	0.3972	16000	2.2262
2.2340	0.4097	16500	2.2211
2.2299	0.4221	17000	2.2240
2.2203	0.4345	17500	2.2275
2.2294	0.4469	18000	2.2207
2.2140	0.4593	18500	2.2183
2.2174	0.4717	19000	2.2211
2.2180	0.4841	19500	2.2171
2.2164	0.4965	20000	2.2175
2.2127	0.5090	20500	2.2183
2.2223	0.5214	21000	2.2135
2.2097	0.5338	21500	2.2140
2.2127	0.5462	22000	2.2115
2.2099	0.5586	22500	2.2131
2.2100	0.5710	23000	2.2119
2.2121	0.5834	23500	2.2117
2.2125	0.5959	24000	2.2100
2.2139	0.6083	24500	2.2090
2.2176	0.6207	25000	2.2088
2.2009	0.6331	25500	2.2059
2.2018	0.6455	26000	2.2085
2.2010	0.6579	26500	2.2056
2.2048	0.6703	27000	2.2064
2.2001	0.6828	27500	2.2058
2.2066	0.6952	28000	2.2020
2.2014	0.7076	28500	2.2042
2.2040	0.7200	29000	2.1990
2.1964	0.7324	29500	2.2009
2.1971	0.7448	30000	2.2022
2.2017	0.7572	30500	2.2016
2.1990	0.7697	31000	2.1978
2.1970	0.7821	31500	2.2035
2.1924	0.7945	32000	2.1991
2.1985	0.8069	32500	2.1962
2.2035	0.8193	33000	2.2007
2.2020	0.8317	33500	2.1956
2.2070	0.8441	34000	2.1966
2.2005	0.8565	34500	2.1990
2.2004	0.8690	35000	2.1980
2.1900	0.8814	35500	2.1963
2.1900	0.8938	36000	2.1977
2.2011	0.9062	36500	2.1937
2.1723	0.9186	37000	2.1881
2.1633	0.9310	37500	2.1633
2.1372	0.9434	38000	2.1396
2.1128	0.9559	38500	2.1079
2.0805	0.9683	39000	2.0801
2.0601	0.9807	39500	2.0580
2.0436	0.9931	40000	2.0474
2.0403	1.0	40278	2.0465

Framework versions

Transformers 5.14.1
Pytorch 2.13.0+cu130
Datasets 5.0.0
Tokenizers 0.22.2

Downloads last month: 610

Safetensors

Model size

521k params

Tensor type

F32