|
--- |
|
license: llama2 |
|
train: false |
|
inference: false |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
## Llama-2-7b-hf-4bit_g64-HQQ |
|
This a version of the LLama2-7B model quantized to 4-bit via Half-Quadratic Quantization (HQQ): https://mobiusml.github.io/hqq/ |
|
|
|
To run the model, install the HQQ library from https://github.com/mobiusml/hqq/tree/main/code and load it as follows: |
|
``` Python |
|
from hqq.models.llama import LlamaHQQ |
|
import transformers |
|
|
|
model_id = 'mobiuslabsgmbh/Llama-2-7b-hf-4bit_g64-HQQ' |
|
#Load the tokenizer |
|
tokenizer = transformers.AutoTokenizer.from_pretrained(model_id) |
|
#Load the model |
|
model = LlamaHQQ.from_quantized(model_id) |
|
``` |
|
|
|
You can then use the model for text generation or to reproduce the benchmark numbers. |
|
|
|
*Limitations*: |
|
-Only supports a single GPU runtime |
|
-The model is not compatible with HuggingFace's PEFT. |
|
|
|
|
|
|
|
|
|
|
