89 lines
3.8 KiB
Markdown
89 lines
3.8 KiB
Markdown
|
---
|
||
|
|
license: apache-2.0
|
||
|
|
base_model: allenai/OLMo-2-1124-7B-Instruct
|
||
|
|
tags:
|
||
|
|
- gguf
|
||
|
|
- quantized
|
||
|
|
- olmo
|
||
|
|
- allenai
|
||
|
|
- text-generation
|
||
|
|
- fully-open
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
---
|
||
|
|
|
||
|
|
# OLMo-2-7B-Instruct — GGUF Quants
|
||
|
|
|
||
|
|
Quantized GGUF versions of [allenai/OLMo-2-1124-7B-Instruct](https://huggingface.co/allenai/OLMo-2-1124-7B-Instruct) — the Allen Institute for AI's fully open 7B language model. OLMo-2 is released under Apache 2.0 with **full training data, code, intermediate checkpoints, and evaluation data** all publicly available — the most transparent large language model at this scale.
|
||
|
|
|
||
|
|
## Available Files
|
||
|
|
|
||
|
|
| File | Quant | Size | Use Case |
|
||
|
|
|------|-------|------|----------|
|
||
|
|
| `OLMo-2-7B-Instruct-Q8_0.gguf` | Q8_0 | ~7.4GB | Maximum quality |
|
||
|
|
| `OLMo-2-7B-Instruct-Q6_K.gguf` | Q6_K | ~5.7GB | Near-lossless |
|
||
|
|
| `OLMo-2-7B-Instruct-Q5_K_M.gguf` | Q5_K_M | ~5.0GB | High quality |
|
||
|
|
| `OLMo-2-7B-Instruct-Q4_K_M.gguf` | Q4_K_M | ~4.2GB | **Recommended default** |
|
||
|
|
| `OLMo-2-7B-Instruct-Q3_K_M.gguf` | Q3_K_M | ~3.5GB | Low VRAM |
|
||
|
|
| `OLMo-2-7B-Instruct-IQ4_XS.gguf` | IQ4_XS | ~3.8GB | Imatrix 4-bit |
|
||
|
|
| `OLMo-2-7B-Instruct-IQ3_XXS.gguf` | IQ3_XXS | ~2.7GB | Imatrix 3-bit |
|
||
|
|
| `OLMo-2-7B-Instruct-IQ2_M.gguf` | IQ2_M | ~2.4GB | Imatrix 2-bit |
|
||
|
|
| `OLMo-2-7B-Instruct-IQ1_S.gguf` | IQ1_S | ~1.6GB | Extreme compression |
|
||
|
|
| `OLMo-2-7B-Instruct-fp16.gguf` | FP16 | ~14.0GB | Full precision |
|
||
|
|
| `imatrix.dat` | — | — | Importance matrix |
|
||
|
|
|
||
|
|
## Usage
|
||
|
|
|
||
|
|
```bash
|
||
|
|
# llama.cpp
|
||
|
|
./llama-cli -m OLMo-2-7B-Instruct-Q4_K_M.gguf \
|
||
|
|
--ctx-size 4096 -n 512 \
|
||
|
|
-p "<|endoftext|><|user|>\nHello!<|assistant|>\n"
|
||
|
|
|
||
|
|
# Ollama
|
||
|
|
ollama run hf.co/DuoNeural/OLMo-2-7B-Instruct-GGUF:Q4_K_M
|
||
|
|
```
|
||
|
|
|
||
|
|
## About OLMo-2-7B
|
||
|
|
|
||
|
|
- **Parameters**: 7B
|
||
|
|
- **Context**: 4096 tokens
|
||
|
|
- **Architecture**: Modified decoder-only transformer (RoPE, SwiGLU, QK-norm)
|
||
|
|
- **Training Data**: Dolmino Mix 1124 — fully open dataset
|
||
|
|
- **License**: Apache 2.0 (commercial use OK)
|
||
|
|
- **Unique value**: All artifacts open — weights, data, code, evals, checkpoints
|
||
|
|
|
||
|
|
OLMo-2 is the gold standard for research reproducibility in the 7B class. If you need a model where you can trace every decision back to first principles, this is it.
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
*Quantized by DuoNeural using llama.cpp on RTX 5090.*
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
## DuoNeural
|
||
|
|
|
||
|
|
**DuoNeural** is an open AI research lab — human + AI in collaboration.
|
||
|
|
|
||
|
|
| Platform | Link |
|
||
|
|
|----------|------|
|
||
|
|
| HuggingFace | [huggingface.co/DuoNeural](https://huggingface.co/DuoNeural) |
|
||
|
|
| Website | [duoneural.com](https://duoneural.com) |
|
||
|
|
| GitHub | [github.com/DuoNeural](https://github.com/DuoNeural) |
|
||
|
|
| X / Twitter | [@DuoNeural](https://x.com/DuoNeural) |
|
||
|
|
| Email | duoneural@proton.me |
|
||
|
|
| Newsletter | [duoneural.beehiiv.com](https://duoneural.beehiiv.com) |
|
||
|
|
| Support | [buymeacoffee.com/duoneural](https://buymeacoffee.com/duoneural) |
|
||
|
|
|
||
|
|
### DuoNeural Research Publications
|
||
|
|
|
||
|
|
| Title | DOI |
|
||
|
|
|-------|-----|
|
||
|
|
| [Nano-CTM: Ternary Continuous Thought Machines with Thought-Space Self-Prediction for Efficient Iterative Reasoning](https://doi.org/10.5281/zenodo.19775622) | [10.5281/zenodo.19775622](https://doi.org/10.5281/zenodo.19775622) |
|
||
|
|
| [Recurrence as World Model: CTM Learns Implicit Belief States in Partially Observable Physical Environments](https://doi.org/10.5281/zenodo.19810620) | [10.5281/zenodo.19810620](https://doi.org/10.5281/zenodo.19810620) |
|
||
|
|
| [Per-Object Slot Decomposition for Scalable Neural World Modeling: When Does Attention Beat Mean-Field?](https://doi.org/10.5281/zenodo.19846804) | [10.5281/zenodo.19846804](https://doi.org/10.5281/zenodo.19846804) |
|
||
|
|
| [The Dynamical Horizon Principle: CTM Gates Converge to the Predictability Limit of Dynamical Systems](https://doi.org/10.5281/zenodo.19952612) | [10.5281/zenodo.19952612](https://doi.org/10.5281/zenodo.19952612) |
|
||
|
|
|
||
|
|
*Open access, CC BY 4.0. Authored by Archon, Jesse Caldwell, Aura — DuoNeural.*
|