AQLM+PV
Collection
Official AQLM quantizations for "PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression": https://arxiv.org/abs/2405.14852
•
25 items
•
Updated
•
20
Official AQLM quantization of meta-llama/Llama-3.2-1B finetuned with PV-Tuning.
For this quantization, we used 2 codebooks of 8 bits and groupsize of 8.
Results:
| Model | Quantization | MMLU (5-shot) | ArcC | ArcE | Hellaswag | PiQA | Winogrande | Model size, Gb |
|---|---|---|---|---|---|---|---|---|
| meta-llama/Llama-3.2-1B-Instruct | fp16 | 0.4551 | 0.3584 | 0.6856 | 0.4520 | 0.7432 | 0.5951 | 2.5 |
| 2x8g8 | 0.2886 | 0.2918 | 0.6065 | 0.3990 | 0.7057 | 0.5746 | 0.8 |
- Downloads last month
- 192
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
the model is not deployed on the HF Inference API.
Model tree for ISTA-DASLab/Llama-3.2-1B-Instruct-AQLM-PV-2Bit-2x8
Base model
meta-llama/Llama-3.2-1B-Instruct