Instructions to use sapbot/gemma-3n-4b-it-distill-smollm2-360m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sapbot/gemma-3n-4b-it-distill-smollm2-360m with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/SmolLM2-360M-Instruct")
model = PeftModel.from_pretrained(base_model, "sapbot/gemma-3n-4b-it-distill-smollm2-360m")

Transformers

How to use sapbot/gemma-3n-4b-it-distill-smollm2-360m with Transformers:

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

pipe = pipeline("text-generation", model="sapbot/gemma-3n-4b-it-distill-smollm2-360m")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("sapbot/gemma-3n-4b-it-distill-smollm2-360m", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use sapbot/gemma-3n-4b-it-distill-smollm2-360m with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sapbot/gemma-3n-4b-it-distill-smollm2-360m"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sapbot/gemma-3n-4b-it-distill-smollm2-360m",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sapbot/gemma-3n-4b-it-distill-smollm2-360m

SGLang

How to use sapbot/gemma-3n-4b-it-distill-smollm2-360m 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 "sapbot/gemma-3n-4b-it-distill-smollm2-360m" \
    --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": "sapbot/gemma-3n-4b-it-distill-smollm2-360m",
		"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 "sapbot/gemma-3n-4b-it-distill-smollm2-360m" \
        --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": "sapbot/gemma-3n-4b-it-distill-smollm2-360m",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use sapbot/gemma-3n-4b-it-distill-smollm2-360m with Docker Model Runner:
```
docker model run hf.co/sapbot/gemma-3n-4b-it-distill-smollm2-360m
```

gemma-3n-4b-it-distill-smollm2-360m / running_log.txt

sapbot

Upload 15 files

64b84be verified about 1 month ago

raw

history blame contribute delete

13.6 kB

	[INFO\|2026-05-08 20:51:45] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 20:51:45] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 20:51:47] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 20:51:47] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 20:51:47] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 20:51:47] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 20:51:49] logging.py:144 >> Add <\|im_end\|> to stop words.
	[INFO\|2026-05-08 20:51:49] logging.py:144 >> Loading dataset sapbot/gemma-3n-4b-it-423x...
	[INFO\|2026-05-08 20:51:51] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 20:51:51] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 20:51:51] logging.py:144 >> KV cache is disabled during training.
	[INFO\|2026-05-08 20:51:52] modeling_utils.py:710 >> loading weights file model.safetensors from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/model.safetensors
	[INFO\|2026-05-08 20:51:52] modeling_utils.py:779 >> Will use dtype=torch.bfloat16 as defined in model's config object
	[INFO\|2026-05-08 20:51:52] configuration_utils.py:1014 >> Generate config GenerationConfig {
	"bos_token_id": 1,
	"eos_token_id": 2,
	"output_attentions": false,
	"output_hidden_states": false,
	"pad_token_id": 49152,
	"use_cache": false
	}

	[INFO\|2026-05-08 20:51:53] configuration_utils.py:967 >> loading configuration file generation_config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/generation_config.json
	[INFO\|2026-05-08 20:51:53] configuration_utils.py:1014 >> Generate config GenerationConfig {
	"bos_token_id": 1,
	"eos_token_id": 2,
	"max_length": 8192,
	"pad_token_id": 49152
	}

	[INFO\|2026-05-08 20:51:54] dynamic_module_utils.py:406 >> Could not locate the custom_generate/generate.py inside unsloth/SmolLM2-360M-Instruct.
	[INFO\|2026-05-08 20:51:54] logging.py:144 >> Gradient checkpointing enabled.
	[INFO\|2026-05-08 20:51:54] logging.py:144 >> Using torch SDPA for faster training and inference.
	[INFO\|2026-05-08 20:51:54] logging.py:144 >> Upcasting trainable params to float32.
	[INFO\|2026-05-08 20:51:54] logging.py:144 >> Fine-tuning method: LoRA
	[INFO\|2026-05-08 20:51:54] logging.py:144 >> Found linear modules: o_proj,q_proj,down_proj,v_proj,k_proj,up_proj,gate_proj
	[INFO\|2026-05-08 20:51:56] logging.py:144 >> trainable params: 4,341,760 \|\| all params: 366,163,840 \|\| trainable%: 1.1857
	[INFO\|2026-05-08 20:51:56] trainer.py:1587 >> *** Running training ***
	[INFO\|2026-05-08 20:51:56] trainer.py:1588 >> Num examples = 423
	[INFO\|2026-05-08 20:51:56] trainer.py:1589 >> Num Epochs = 3
	[INFO\|2026-05-08 20:51:56] trainer.py:1590 >> Instantaneous batch size per device = 2
	[INFO\|2026-05-08 20:51:56] trainer.py:1593 >> Total train batch size (w. parallel, distributed & accumulation) = 16
	[INFO\|2026-05-08 20:51:56] trainer.py:1594 >> Gradient Accumulation steps = 8
	[INFO\|2026-05-08 20:51:56] trainer.py:1595 >> Total optimization steps = 81
	[INFO\|2026-05-08 20:51:56] trainer.py:1596 >> Number of trainable parameters = 4,341,760
	[INFO\|2026-05-08 20:53:35] logging.py:144 >> {'loss': 4.6997, 'learning_rate': 4.9981e-05, 'epoch': 0.19, 'throughput': 933.89}
	[INFO\|2026-05-08 20:55:17] logging.py:144 >> {'loss': 10.7092, 'learning_rate': 4.9981e-05, 'epoch': 0.38, 'throughput': 954.01}
	[INFO\|2026-05-08 20:56:59] logging.py:144 >> {'loss': 11.7926, 'learning_rate': 4.9981e-05, 'epoch': 0.57, 'throughput': 958.84}
	[INFO\|2026-05-08 20:58:39] logging.py:144 >> {'loss': 13.5184, 'learning_rate': 4.9925e-05, 'epoch': 0.75, 'throughput': 956.42}
	[INFO\|2026-05-08 21:00:25] logging.py:144 >> {'loss': 20.1784, 'learning_rate': 4.9925e-05, 'epoch': 0.94, 'throughput': 959.60}
	[INFO\|2026-05-08 21:02:09] logging.py:144 >> {'loss': 7.2012, 'learning_rate': 4.9831e-05, 'epoch': 1.11, 'throughput': 953.38}
	[INFO\|2026-05-08 21:03:58] logging.py:144 >> {'loss': 16.0000, 'learning_rate': 4.9831e-05, 'epoch': 1.30, 'throughput': 952.43}
	[INFO\|2026-05-08 21:05:50] logging.py:144 >> {'loss': 9.6828, 'learning_rate': 4.9831e-05, 'epoch': 1.49, 'throughput': 951.69}
	[INFO\|2026-05-08 21:07:46] logging.py:144 >> {'loss': 6.1018, 'learning_rate': 4.9831e-05, 'epoch': 1.68, 'throughput': 950.28}
	[INFO\|2026-05-08 21:09:18] logging.py:144 >> {'loss': 10.2955, 'learning_rate': 4.9700e-05, 'epoch': 1.87, 'throughput': 953.07}
	[INFO\|2026-05-08 21:11:03] logging.py:144 >> {'loss': 4.0894, 'learning_rate': 4.9326e-05, 'epoch': 2.04, 'throughput': 951.55}
	[INFO\|2026-05-08 21:12:46] logging.py:144 >> {'loss': 26.6340, 'learning_rate': 4.9326e-05, 'epoch': 2.23, 'throughput': 952.42}
	[INFO\|2026-05-08 21:14:35] logging.py:144 >> {'loss': 5.8759, 'learning_rate': 4.9084e-05, 'epoch': 2.42, 'throughput': 950.59}
	[INFO\|2026-05-08 21:16:32] logging.py:144 >> {'loss': 4.9436, 'learning_rate': 4.9084e-05, 'epoch': 2.60, 'throughput': 947.64}
	[INFO\|2026-05-08 21:18:21] logging.py:144 >> {'loss': 11.3494, 'learning_rate': 4.9084e-05, 'epoch': 2.79, 'throughput': 948.32}
	[INFO\|2026-05-08 21:19:45] logging.py:144 >> {'loss': 3.3956, 'learning_rate': 4.8492e-05, 'epoch': 2.98, 'throughput': 950.90}
	[INFO\|2026-05-08 21:19:53] trainer.py:3797 >> Saving model checkpoint to saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08/checkpoint-81
	[INFO\|2026-05-08 21:19:54] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 21:19:54] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 21:19:54] tokenization_utils_base.py:3224 >> chat template saved in saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08/checkpoint-81/chat_template.jinja
	[INFO\|2026-05-08 21:19:54] tokenization_utils_base.py:2078 >> tokenizer config file saved in saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08/checkpoint-81/tokenizer_config.json
	[INFO\|2026-05-08 21:19:54] trainer.py:1863 >>

	Training completed. Do not forget to share your model on huggingface.co/models =)


	[INFO\|2026-05-08 21:19:54] trainer.py:3797 >> Saving model checkpoint to saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08
	[INFO\|2026-05-08 21:19:55] configuration_utils.py:670 >> loading configuration file config.json from cache at /home/sapbot/.cache/huggingface/hub/models--unsloth--SmolLM2-360M-Instruct/snapshots/7c6bf9a2faef35209646cd7f1569fae2b2394a00/config.json
	[INFO\|2026-05-08 21:19:55] configuration_utils.py:742 >> Model config LlamaConfig {
	"architectures": [
	"LlamaForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 1,
	"dtype": "bfloat16",
	"eos_token_id": 2,
	"head_dim": 64,
	"hidden_act": "silu",
	"hidden_size": 960,
	"initializer_range": 0.02,
	"intermediate_size": 2560,
	"is_llama_config": true,
	"max_position_embeddings": 8192,
	"mlp_bias": false,
	"model_type": "llama",
	"num_attention_heads": 15,
	"num_hidden_layers": 32,
	"num_key_value_heads": 5,
	"pad_token_id": 49152,
	"pretraining_tp": 1,
	"rms_norm_eps": 1e-05,
	"rope_interleaved": false,
	"rope_parameters": {
	"rope_theta": 100000,
	"rope_type": "default"
	},
	"tie_word_embeddings": true,
	"transformers.js_config": {
	"kv_cache_dtype": {
	"fp16": "float16",
	"q4f16": "float16"
	}
	},
	"transformers_version": "5.2.0",
	"use_cache": true,
	"vocab_size": 49153
	}

	[INFO\|2026-05-08 21:19:55] tokenization_utils_base.py:3224 >> chat template saved in saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08/chat_template.jinja
	[INFO\|2026-05-08 21:19:55] tokenization_utils_base.py:2078 >> tokenizer config file saved in saves/SmolLM-360M-Instruct/lora/train_2026-05-08-20-46-08/tokenizer_config.json
	[WARNING\|2026-05-08 21:19:55] logging.py:149 >> No metric eval_loss to plot.
	[WARNING\|2026-05-08 21:19:55] logging.py:149 >> No metric eval_accuracy to plot.
	[INFO\|2026-05-08 21:19:55] modelcard.py:266 >> Dropping the following result as it does not have all the necessary fields:
	{'task': {'name': 'Causal Language Modeling', 'type': 'text-generation'}}