Instructions to use mudkip02/llama3.2-1b-mmqa-full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mudkip02/llama3.2-1b-mmqa-full with Transformers:

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

pipe = pipeline("text-generation", model="mudkip02/llama3.2-1b-mmqa-full")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("mudkip02/llama3.2-1b-mmqa-full")
model = AutoModelForCausalLM.from_pretrained("mudkip02/llama3.2-1b-mmqa-full")
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 mudkip02/llama3.2-1b-mmqa-full with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mudkip02/llama3.2-1b-mmqa-full"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mudkip02/llama3.2-1b-mmqa-full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mudkip02/llama3.2-1b-mmqa-full

SGLang

How to use mudkip02/llama3.2-1b-mmqa-full 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 "mudkip02/llama3.2-1b-mmqa-full" \
    --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": "mudkip02/llama3.2-1b-mmqa-full",
		"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 "mudkip02/llama3.2-1b-mmqa-full" \
        --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": "mudkip02/llama3.2-1b-mmqa-full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mudkip02/llama3.2-1b-mmqa-full with Docker Model Runner:
```
docker model run hf.co/mudkip02/llama3.2-1b-mmqa-full
```

llama3.2-1b-mmqa-full-llamafactory

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct on the meta_math_qa dataset. It achieves the following results on the evaluation set:

Loss: 0.1648

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: 2e-06
train_batch_size: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 0.03
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2472	0.0102	250	0.2769
0.2498	0.0203	500	0.2437
0.2334	0.0305	750	0.2324
0.2390	0.0406	1000	0.2250
0.2166	0.0508	1250	0.2208
0.2573	0.0609	1500	0.2164
0.2134	0.0711	1750	0.2134
0.2163	0.0812	2000	0.2108
0.1992	0.0914	2250	0.2085
0.2166	0.1015	2500	0.2057
0.1944	0.1117	2750	0.2051
0.2081	0.1218	3000	0.2040
0.1851	0.1320	3250	0.2026
0.2131	0.1421	3500	0.2024
0.2037	0.1523	3750	0.1993
0.1865	0.1624	4000	0.1973
0.1969	0.1726	4250	0.1968
0.1983	0.1827	4500	0.1952
0.2103	0.1929	4750	0.1940
0.1968	0.2030	5000	0.1941
0.2135	0.2132	5250	0.1933
0.2059	0.2234	5500	0.1927
0.1912	0.2335	5750	0.1912
0.2004	0.2437	6000	0.1902
0.1868	0.2538	6250	0.1898
0.1946	0.2640	6500	0.1888
0.1780	0.2741	6750	0.1879
0.1804	0.2843	7000	0.1874
0.1798	0.2944	7250	0.1866
0.1926	0.3046	7500	0.1850
0.1829	0.3147	7750	0.1848
0.1854	0.3249	8000	0.1841
0.1703	0.3350	8250	0.1833
0.1715	0.3452	8500	0.1828
0.1811	0.3553	8750	0.1822
0.1876	0.3655	9000	0.1808
0.2054	0.3756	9250	0.1811
0.1994	0.3858	9500	0.1815
0.1866	0.3959	9750	0.1805
0.1925	0.4061	10000	0.1800
0.1797	0.4162	10250	0.1793
0.1613	0.4264	10500	0.1785
0.1702	0.4365	10750	0.1783
0.1755	0.4467	11000	0.1778
0.1735	0.4569	11250	0.1774
0.1668	0.4670	11500	0.1765
0.1786	0.4772	11750	0.1761
0.1808	0.4873	12000	0.1763
0.1774	0.4975	12250	0.1756
0.1796	0.5076	12500	0.1748
0.1744	0.5178	12750	0.1744
0.1658	0.5279	13000	0.1743
0.1853	0.5381	13250	0.1741
0.1702	0.5482	13500	0.1734
0.1688	0.5584	13750	0.1727
0.1684	0.5685	14000	0.1722
0.1654	0.5787	14250	0.1719
0.1595	0.5888	14500	0.1713
0.1817	0.5990	14750	0.1712
0.1600	0.6091	15000	0.1706
0.1549	0.6193	15250	0.1702
0.1763	0.6294	15500	0.1699
0.1390	0.6396	15750	0.1698
0.1544	0.6497	16000	0.1696
0.1664	0.6599	16250	0.1691
0.1824	0.6701	16500	0.1687
0.1641	0.6802	16750	0.1682
0.1621	0.6904	17000	0.1682
0.1643	0.7005	17250	0.1681
0.1614	0.7107	17500	0.1676
0.1699	0.7208	17750	0.1673
0.1648	0.7310	18000	0.1672
0.1942	0.7411	18250	0.1669
0.1739	0.7513	18500	0.1667
0.1734	0.7614	18750	0.1666
0.1672	0.7716	19000	0.1663
0.1683	0.7817	19250	0.1660
0.1631	0.7919	19500	0.1659
0.1689	0.8020	19750	0.1658
0.1493	0.8122	20000	0.1657
0.1673	0.8223	20250	0.1658
0.1615	0.8325	20500	0.1657
0.1609	0.8426	20750	0.1656
0.1653	0.8528	21000	0.1653
0.1628	0.8629	21250	0.1653
0.1623	0.8731	21500	0.1652
0.1540	0.8832	21750	0.1651
0.1551	0.8934	22000	0.1650
0.1609	0.9036	22250	0.1650
0.1761	0.9137	22500	0.1649
0.1707	0.9239	22750	0.1649
0.1862	0.9340	23000	0.1649
0.1691	0.9442	23250	0.1649
0.1435	0.9543	23500	0.1649
0.1680	0.9645	23750	0.1648
0.1594	0.9746	24000	0.1649
0.1743	0.9848	24250	0.1648
0.1751	0.9949	24500	0.1648
0.1835	1.0	24625	0.1648

Framework versions

Transformers 5.6.0
Pytorch 2.8.0+cu126
Datasets 4.0.0
Tokenizers 0.22.2

Downloads last month: 26

Safetensors

Model size

1B params

Tensor type

F32

Model tree for mudkip02/llama3.2-1b-mmqa-full

Base model

meta-llama/Llama-3.2-1B-Instruct

Finetuned

(1769)

this model