| ---
|
| language:
|
| - en
|
| tags:
|
| - gpt2
|
| - scaling-study
|
| - benchmarking
|
| - banterhearts
|
| pipeline_tag: text-generation
|
| library_name: transformers
|
| license: mit
|
| ---
|
|
|
| # Tiny GPT-2 (4.6M)
|
|
|
| Custom-trained GPT-2 checkpoint with deliberate depth-width configuration for inference benchmarking research.
|
|
|
| Created as part of the [Banterhearts research program](https://github.com/Sahil170595/Banterhearts) investigating benchmarking integrity for local LLM inference.
|
|
|
| | | |
|
| |---|---|
|
| | **Architecture** | GPT2LMHeadModel (MHA) |
|
| | **Parameters** | 4.6M |
|
| | **Config** | n_embd=2, n_head=2, n_layer=2 |
|
| | **Context length** | 1,024 tokens |
|
| | **Precision** | FP32 |
|
| | **Model size** | 2.4 MB |
|
| | **Vocab size** | 50,257 |
|
|
|
| ## Purpose
|
|
|
| Environment validation and weight parity checks.
|
|
|
| These checkpoints are not general-purpose language models. They are deliberately sized scaling-study artifacts designed to isolate the effect of model depth vs width on GPU inference latency. The key finding: in the small-model GPU regime, **layer depth** (not parameter count) dominates latency, producing inversions where a 5M-parameter model can be 3.6x slower than a 25M-parameter model.
|
|
|
| ## Source Technical Reports
|
|
|
| Used in: TR126, TR147
|
|
|
| | TR | Role |
|
| |---|---|
|
| | TR117 | Original cross-backend benchmark matrix (7 backends, 4 model groups) |
|
| | TR126 | Linux/Triton compiler validation with phase-separated measurement |
|
| | TR147 | Second-regime portability validation on RTX 6000 Ada |
|
|
|
| ## Design Rationale
|
|
|
| The GPT-2 family (25M, 50M, 100M) uses a 2x3 factorial design:
|
|
|
| | Model | n_embd | n_layer | n_inner | Params | Design role |
|
| |---|---|---|---|---|---|
|
| | gpt2-25m | 384 | 3 | 1,536 | 25M | Shallow, narrow |
|
| | gpt2-50m | 512 | 8 | 2,048 | 50M | Deep, medium width |
|
| | gpt2-100m | 768 | 8 | 3,072 | 100M | Deep, wide |
|
|
|
| All models use **2 attention heads** (MHA, not GQA) to isolate architecture effects from attention-group structure. Dropout is set to 0.0 for deterministic inference measurement.
|
|
|
| ## Usage
|
|
|
| ```python
|
| from transformers import AutoModelForCausalLM, AutoTokenizer
|
|
|
| model = AutoModelForCausalLM.from_pretrained("Crusadersk/tiny-gpt2")
|
| tokenizer = AutoTokenizer.from_pretrained("Crusadersk/tiny-gpt2")
|
|
|
| inputs = tokenizer("Hello", return_tensors="pt")
|
| outputs = model.generate(**inputs, max_new_tokens=32, do_sample=False)
|
| print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| ```
|
|
|
| ## Compatibility
|
|
|
| | Framework | Supported |
|
| |---|---|
|
| | Transformers | Yes |
|
| | torch.compile (Inductor) | Yes |
|
| | Ollama | No (not GGUF format) |
|
| | vLLM | Yes |
|
|
|
| ## Citation
|
|
|
| ```bibtex
|
| @misc{banterhearts2026tinygpt2,
|
| title = {Custom GPT-2 Scaling Checkpoint (4.6M) for Inference Benchmarking Research},
|
| author = {Kadadekar, Sahil},
|
| year = {2026},
|
| url = {https://huggingface.co/Crusadersk/tiny-gpt2},
|
| note = {Part of the Banterhearts research program. NeurIPS 2026 submission.}
|
| }
|
| ```
|
|
|
| ## Acknowledgments
|
|
|
| This work is part of a 40-TR research program on consumer LLM deployment safety, conducted independently as pre-doctoral research. Full program details at [github.com/Sahil170595/Banterhearts](https://github.com/Sahil170595/Banterhearts).
|
|
|
