Instructions to use User01110/tinyLM-8M-exp with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use User01110/tinyLM-8M-exp with Transformers:

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

pipe = pipeline("text-generation", model="User01110/tinyLM-8M-exp", trust_remote_code=True)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("User01110/tinyLM-8M-exp", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("User01110/tinyLM-8M-exp", trust_remote_code=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use User01110/tinyLM-8M-exp with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "User01110/tinyLM-8M-exp"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "User01110/tinyLM-8M-exp",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/User01110/tinyLM-8M-exp

SGLang

How to use User01110/tinyLM-8M-exp 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 "User01110/tinyLM-8M-exp" \
    --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": "User01110/tinyLM-8M-exp",
		"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 "User01110/tinyLM-8M-exp" \
        --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": "User01110/tinyLM-8M-exp",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use User01110/tinyLM-8M-exp with Docker Model Runner:
```
docker model run hf.co/User01110/tinyLM-8M-exp
```

tinyLM-8M-exp

Tiny 5M-class parameter Qwen3-config causal LM with math-only novelty-gated GQA.

Architecture

Item	Value
Config type	`qwen3`
Parameters	8.919M
Layers	10
Hidden size	256
MLP size	768
Query heads	8
KV heads	4
Head dim	32
RoPE theta	2500
Tied embeddings	yes

Attention	Value
Type	GQA
Novelty gate	math-only element-wise RMS-normalized abs-delta
Gate floor	0.05

Training

Item	Value
Tokenizer	`AxiomicLabs/GPT-S2-5M`
Sequence length	512
Microbatch size	512
Gradient accumulation	4
Effective batch size	2048
Steps	20,000
Validation cadence	every 1,000 steps
Raw MC eval cadence	every 2,000 steps on ARC-Easy, ARC-Challenge, PIQA, HellaSwag
LR schedule	warmup, cosine to min by 10,000, hold min to 15,000, cosine tail to zero by 20,000
Optimizer	Muon for middle 2D weights, AdamW for the rest
Special-token policy	BOS/EOS are document-level; `<

Dataset	Share	Config
`HuggingFaceFW/fineweb-edu`	60.0%	`sample-100BT`
`HuggingFaceTB/smollm-corpus`	30.0%	`cosmopedia-v2` only
`epfml/FineWeb-HQ`	10.0%	`default`

Validation

Metric	Value
Dataset	`Salesforce/wikitext`, `wikitext-103-raw-v1`, validation
Context / stride	512 / 256
Loss	3.1546
Perplexity	23.44
UTF-8 BPB	1.4433
Scored tokens	365,258
UTF-8 bytes	1,151,766

Load And Generate

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "User01110/tinyLM-8M-exp"
tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True)

inputs = tokenizer("Once upon a time", return_tensors="pt").to(model.device)
print(inputs.input_ids[0][:2].tolist())  # [<|im_start|>, <|bos|>]

with torch.no_grad():
    output = model.generate(**inputs, max_new_tokens=128, do_sample=True, temperature=0.7, top_k=40)

print(tokenizer.decode(output[0], skip_special_tokens=True))

This repo uses a native Qwen3Config plus remote model code for the math-only novelty-gated attention block.

Downloads last month: -

Safetensors

Model size

8.92M params

Tensor type

F32

Model tree for User01110/tinyLM-8M-exp

Base model

Qwen/Qwen3-0.6B-Base

Finetuned

Qwen/Qwen3-0.6B

Finetuned

(1006)

this model