PY

Pythia 2.8b

PythiaGGUF by mradermacherUpdated 2024-09-08
Download on Hugging Face
Parameters2.78B
Context2K
Layers32
Architecturegptneox
Smallest quant0.7 GB
Quick answer

Pythia 2.8b needs about 5.0 GB of memory at Q4_K_M (a 1.8 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 24 quants, from 0.7 GB (IQ1_S) to 2.3 GB (Q6_K).

Pythia 2.8b – 2.78B‑parameter GPT-NeoX model

Pythia 2.8b is a 2.78B‑parameter transformer trained on the Pile. It is part of the Pythia Scaling Suite and is released under Apache‑2.0. The model supports 2048‑token context and is available in multiple GGUF quantizations.

Good at

  • Large‑context language modeling
  • Research‑grade interpretability
  • Stable inference on mid‑range GPUs

Limits

  • No vision support
  • Requires 2048‑token context
  • License is Apache‑2.0

Best for

  • Academic research on LLMs
  • Fine‑tuning on custom corpora
  • Benchmarking quantization methods

Who it’s for. Researchers and developers who need a well‑documented, open‑source language model for experimentation. The model is suitable for those who prefer a transformer with a moderate parameter count and a variety of quantization options.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S0.7 GB 1.8 GB
Fits Get file →
IQ1_M0.7 GB 1.9 GB
Fits Get file →
IQ2_XXS0.8 GB 2.0 GB
Fits Get file →
IQ2_XS0.9 GB 2.0 GB
Fits Get file →
IQ2_S0.9 GB 2.1 GB
Fits Get file →
IQ2_M1.0 GB 2.2 GB
Fits Get file →
Q2_K1.1 GB 2.3 GB
Fits Get file →
IQ3_XXS1.1 GB 2.3 GB
Fits Get file →
IQ3_XS1.2 GB 2.4 GB
Fits Get file →
IQ3_S1.2 GB 2.4 GB
Fits Get file →
Q3_K_S1.2 GB 2.4 GB
Fits Get file →
IQ3_M1.4 GB 2.5 GB
Fits Get file →
Q3_K_M1.5 GB 2.6 GB
Fits Get file →
IQ4_XS1.5 GB 2.7 GB
Fits Get file →
Q4_01.6 GB 2.8 GB
Fits Get file →
Q4_01.6 GB 2.8 GB
Fits Get file →
Q4_01.6 GB 2.8 GB
Fits Get file →
Q3_K_L1.6 GB 2.8 GB
Fits Get file →
Q4_01.6 GB 2.8 GB
Fits Get file →
Q4_K_S1.6 GB 2.8 GB
Fits Get file →
Q4_K_M1.8 GB 3.0 GB
Fits Get file →
Q5_K_S1.9 GB 3.1 GB
Fits Get file →
Q5_K_M2.1 GB 3.2 GB
Fits Get file →
Q6_K2.3 GB 3.5 GB
Fits Get file →

For 8 GB VRAM, use IQ1_S (0.66 GB) or IQ1_M (0.71 GB). If 12 GB VRAM is available, Q4_0 (1.6 GB) or Q4_K_S (1.61 GB) provide a good balance of size and performance. For 16 GB VRAM, Q5_K_M (2.07 GB) or Q6_K (2.28 GB) are recommended.

What GPU do you need for Pythia 2.8b?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is IQ2_XXS (0.8 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is Q6_K (2.3 GB file, about 5.5 GB in use).

Which Pythia 2.8b quant should you choose?

Q4_K_M is the usual starting point for Pythia 2.8b: it balances file size and quality well. Here it is a 1.8 GB file.

If memory is tight, the smallest file is IQ1_S at 0.7 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegptneox
Parameters2.78B
Layers32
Trained context2,048 tokens
KV cache (fp16)320 KB per token

Source

Quantized bymradermacher
FamilyPythia
Licenseapache-2.0
Downloads (HF)1,546
Last updated2024-09-08

Frequently asked questions about Pythia 2.8b

How much VRAM does Pythia 2.8b need?

About 5.0 GB at Q4_K_M with an 8K context window. That is the 1.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Pythia 2.8b on an 8 GB GPU?

Yes, with Q6_K: the 2.3 GB file needs about 5.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Pythia 2.8b GGUF?

It depends on the quant: IQ1_S is 0.7 GB, IQ1_M is 0.7 GB, IQ2_XXS is 0.8 GB, IQ2_XS is 0.9 GB, IQ2_S is 0.9 GB, IQ2_M is 1.0 GB and 18 more.

What context length does Pythia 2.8b support?

The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 2.7 GB at 8K and 0.7 GB at 2K.

Where can I download Pythia 2.8b GGUF?

The files are on Hugging Face in the mradermacher/pythia-2.8b-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Pythia 2.8b use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Pythia 2.8b on 8 GB VRAM?

Yes, using the IQ1_S or IQ1_M GGUF quantizations which require approximately 3.8–3.9 GB of VRAM at 8k context.

Does Pythia 2.8b support vision tasks?

No, the model is a text‑only transformer and does not include vision capabilities.

What license applies to Pythia 2.8b?

The model is released under the Apache‑2.0 license.

Other GGUFs of pythia-2.8b

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Pythia models

All Pythia →

Other tiny models

All tiny models →