Instructions to use kanishka/opt-babylm2-subset-default-1e-3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kanishka/opt-babylm2-subset-default-1e-3 with Transformers:

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

pipe = pipeline("text-generation", model="kanishka/opt-babylm2-subset-default-1e-3")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("kanishka/opt-babylm2-subset-default-1e-3")
model = AutoModelForCausalLM.from_pretrained("kanishka/opt-babylm2-subset-default-1e-3")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use kanishka/opt-babylm2-subset-default-1e-3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kanishka/opt-babylm2-subset-default-1e-3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kanishka/opt-babylm2-subset-default-1e-3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/kanishka/opt-babylm2-subset-default-1e-3

SGLang

How to use kanishka/opt-babylm2-subset-default-1e-3 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 "kanishka/opt-babylm2-subset-default-1e-3" \
    --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": "kanishka/opt-babylm2-subset-default-1e-3",
		"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 "kanishka/opt-babylm2-subset-default-1e-3" \
        --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": "kanishka/opt-babylm2-subset-default-1e-3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use kanishka/opt-babylm2-subset-default-1e-3 with Docker Model Runner:
```
docker model run hf.co/kanishka/opt-babylm2-subset-default-1e-3
```

opt-babylm2-subset-default-1e-3

This model was trained from scratch on the kanishka/babylm2-subset dataset. It achieves the following results on the evaluation set:

Loss: 2.3689
Accuracy: 0.5348

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: 0.001
train_batch_size: 32
eval_batch_size: 64
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 32000
num_epochs: 10.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.5477	1.0	14142	2.7490	0.4855
2.377	2.0	28284	2.5803	0.5045
2.2625	3.0	42426	2.4760	0.5164
2.1715	4.0	56568	2.4206	0.5235
2.0996	5.0	70710	2.3880	0.5278
2.0456	6.0	84852	2.3722	0.5306
1.9983	7.0	98994	2.3592	0.5327
1.9482	8.0	113136	2.3579	0.5338
1.9061	9.0	127278	2.3605	0.5346
1.8692	10.0	141420	2.3689	0.5348

Framework versions

Transformers 4.42.4
Pytorch 2.2.0+cu121
Datasets 2.16.1
Tokenizers 0.19.1

Downloads last month: 6

Safetensors

Model size

97.8M params

Tensor type

F32

Dataset used to train kanishka/opt-babylm2-subset-default-1e-3

Evaluation results

Accuracy on kanishka/babylm2-subset
self-reported

0.535