Nous-Hermes-2-Mixtral-8x7B-DPO-4bit
mlx-community/Nous-Hermes-2-Mixtral-8x7B-DPO-4bit
Uncategorised · mlx
4-bit Open on Hugging Face
Parameters
—
Weights (est.)
—
Mac memory (est.)
—
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Details
- Base model
- mistralai/Mixtral-8x7B-v0.1 (finetune)
- Vendor
- Mistral
- Task
- Uncategorised
- Quantisation
- 4-bit
- Licence
- apache-2.0
- Downloads
- 662 in the last 30 days · 9.6K all time · 19 likes
- Published
- 2024-01-17 · updated 2024-01-17 (2y ago)
- Languages
- en
- Tags
- mixtral, Mixtral, instruct, finetune, chatml, DPO, RLHF, gpt4, synthetic data, distillation
Other MLX builds of mistralai/Mixtral-8x7B-v0.1
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| Nous-Hermes-2-Mixtral-8x7B-DPO-8bit | 8-bit | ≈ 49.6 GB | 64 GB+ | 36 |
| NousHermes-Mixtral-8x7B-Reddit-mlx | — | ≈ 186 GB | 256 GB+ | 107 |
| Nous-Hermes-2-Mixtral-8x7B-DPO-4bit (this one) | 4-bit | — | — | 662 |