metadata
license: apache-2.0
tags:
- moe
train: false
inference: false
pipeline_tag: text-generation
Mixtral-8x7B-v0.1-hf-attn-4bit-moe-2bit-HQQ
This is a version of the Mixtral-8x7B-v0.1 model (https://huggingface.co/mistralai/Mixtral-8x7B-v0.1) quantized with a mix of 4-bit and 2-bit via Half-Quadratic Quantization (HQQ).
More specifically, the attention layers are quantized to 4-bit and the experts are quantized to 2-bit. This simple change yields a huge improvement in perplexity vs the all 2-bit model (4.69 vs. 5.90) for a slight increase in model size (18.2GB vs. 18GB).
This idea was suggest by Artem Eliseev (@lavawolfiee) and Denis Mazur (@dvmazur) in this Github discussion.
Basic Usage
To run the model, install the HQQ library from https://github.com/mobiusml/hqq and use it as follows:
model_id = 'mobiuslabsgmbh/Mixtral-8x7B-v0.1-hf-attn-4bit-moe-2bit-HQQ'
#Load the model
from hqq.engine.hf import HQQModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = HQQModelForCausalLM.from_quantized(model_id)
#Optional
from hqq.core.quantize import *
HQQLinear.set_backend(HQQBackend.PYTORCH_COMPILE)
Quantization
You can reproduce the model using the following quant configs:
from hqq.engine.hf import HQQModelForCausalLM, AutoTokenizer
model_id = "mistralai/Mixtral-8x7B-v0.1"
model = HQQModelForCausalLM.from_pretrained(model_id, use_auth_token=hf_auth, cache_dir=cache_path)
#Quantize params
from hqq.core.quantize import *
attn_prams = BaseQuantizeConfig(nbits=4, group_size=64, quant_zero=True, quant_scale=True)
attn_prams['scale_quant_params']['group_size'] = 256
experts_params = BaseQuantizeConfig(nbits=2, group_size=16, quant_zero=True, quant_scale=True)
quant_config = {}
#Attention
quant_config['self_attn.q_proj'] = attn_prams
quant_config['self_attn.k_proj'] = attn_prams
quant_config['self_attn.v_proj'] = attn_prams
quant_config['self_attn.o_proj'] = attn_prams
#Experts
quant_config['block_sparse_moe.experts.w1'] = experts_params
quant_config['block_sparse_moe.experts.w2'] = experts_params
quant_config['block_sparse_moe.experts.w3'] = experts_params
#Quantize
model.quantize_model(quant_config=quant_config)