Instructions to use project-iconik/basikGPT-1-v1.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use project-iconik/basikGPT-1-v1.1 with Transformers:

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

pipe = pipeline("text-generation", model="project-iconik/basikGPT-1-v1.1")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("project-iconik/basikGPT-1-v1.1")
model = AutoModelForCausalLM.from_pretrained("project-iconik/basikGPT-1-v1.1", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use project-iconik/basikGPT-1-v1.1 with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/project-iconik/basikGPT-1-v1.1

SGLang

How to use project-iconik/basikGPT-1-v1.1 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 "project-iconik/basikGPT-1-v1.1" \
    --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": "project-iconik/basikGPT-1-v1.1",
		"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 "project-iconik/basikGPT-1-v1.1" \
        --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": "project-iconik/basikGPT-1-v1.1",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use project-iconik/basikGPT-1-v1.1 with Docker Model Runner:
```
docker model run hf.co/project-iconik/basikGPT-1-v1.1
```

basikGPT-1 v1.1

124M parameters (GPT-2 Small). v1.1 = 5B training tokens, not 5B parameters.

Continues basikGPT-1 v1.0 (FineWeb-Edu 2.5B) for another 2.5B tokens on FineWeb + OpenWebMath.

from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("project-iconik/basikGPT-1-v1.1")
model = AutoModelForCausalLM.from_pretrained("project-iconik/basikGPT-1-v1.1")

Code and training recipe: github.com/project-iconik/basikGPT.
Whitepaper: EN · JA · KO.

Model details


Architecture	GPT-2 Small decoder-only
Parameters	124,439,808
Context	1024
Vocab	50,257 (GPT-2 BPE / tiktoken `gpt2`)
Lifetime tokens	5,000,003,584
This stage	FineWeb 2.25B + OpenWebMath 0.25B (no shuffle; 9:1 shard cycle)
Resume from	v1.0 / `runs/main_2p5b/step-00038147.pt`
Checkpoint	`runs/cont_5b_mix/step-00076294.pt` (step 76,294)
License (weights)	Apache-2.0

This Hub snapshot is a GPT2LMHeadModel export (safetensors + official GPT-2 tokenizer files). Optimizer state is not included.

Training data

Lifetime mix: FineWeb-Edu 50% + FineWeb 45% + OpenWebMath 5%.

HuggingFaceFW/fineweb-edu — first 2.5B tokens (v1.0).
HuggingFaceFW/fineweb (sample-10BT, revision 9bb295ddab0e05d785b879661af7260fed5140fc) — 2.25B tokens.
open-web-math/open-web-math (revision fde8ef8de2300f5e778f56261843dab89f230815) — 0.25B tokens.

FineWeb / FineWeb-Edu: ODC-By 1.0. OpenWebMath: see the dataset card. Downstream use should respect those licenses and original page licenses.

How it was trained

Resumed from v1.0 (did not retrain from scratch). Continuation LR: rewarm 6e-5 → 3e-4 over 1000 steps relative to origin step 38,147, then cosine to 6e-5 at 5B. Same microbatch recipe as v1.0 (BF16, SDPA, 65,536 tokens/step). --no-shuffle --track-data-index --reset-data-index on first resume. Details: GitHub runs/cont_5b_mix/ and docs/whitepaper.md.

FineWeb-Edu validation CE rose during this stage (about 3.32 → 3.47). That is expected: the val set is still FineWeb-Edu while train leaves Edu.

Evaluation (same protocol as the repo)

Zero-shot English LM suite (english-lm-suite-v1) measured in-repo (not mixed with published paper numbers). Token counts and architectures are not matched across rows.

Model	size	HS	LAMBADA	PIQA	WG	ARC-E	Avg
v1.0	124M	29.40	19.58	61.37	50.51	43.01	40.77
v1.1	124M	28.75	23.05	61.75	50.83	38.51	40.58
openai-community/gpt2	124M	30.37	30.93	62.57	51.62	38.13	42.72
HuggingFaceTB/SmolLM2-135M	135M	42.67	42.97	67.57	51.93	59.43	52.91
EleutherAI/pythia-160m	162M	29.26	11.57	58.32	49.49	34.22	36.57
chance		25	—	50	50	~25	—

WG is acc_raw; other columns report the suite's primary metrics. Avg is the unweighted mean of these five primary metrics. n: HS 10,042 · LAMBADA 5,153 · PIQA 1,838 · WG 1,267 · ARC-E 2,376. Full protocol: benchmarks/REPORT.md. Methods and the full comparison: whitepaper.

Versus v1.0: LAMBADA +3.47pp, ARC-Easy −4.50pp, HellaSwag −0.65pp. The Edu-aligned ARC-Easy lead over official gpt2 is essentially gone. WinoGrande remains chance-level. This release does not claim GSM8K gains from 0.25B math tokens (not in the protocol).

Intended use

Research and education: a small, reproducible GPT-2 Small continued on a documented public mix. Not an instruction model. Not safe for open-ended production chat.

Limitations

124M / 5B tokens: still far from modern 135M models trained on much larger mixes.
English-centric web text plus a small math slice; no instruction or preference tuning.
May reproduce biases and factual errors from web text.

Citation

@software{basikgpt,
  title = {basikGPT},
  author = {basikGPT Contributors},
  url = {https://github.com/project-iconik/basikGPT},
  year = {2026}
}

Downloads last month: 527

Safetensors

Model size

0.1B params

Tensor type

F32

Model tree for project-iconik/basikGPT-1-v1.1

Base model

project-iconik/basikGPT-1-v1.0

Finetuned

(1)

this model

project-iconik
/

basikGPT-1-v1.1