HyperCLOVAX-SEED-Think-14B-GPTQ

Instruction

This repo contains GPTQ model files for HyperCLOVAX-SEED-Think-14B.

HyperCLOVAX-SEED-Think-14B-GPTQ was quantized using gptqmodel v4.0.0, following the guide.

Model Configuration

• Original model: naver-hyperclovax/HyperCLOVAX-SEED-Think-14B
• Quantization: GPTQ with 4-bit group-wise weight-only quantization (W4A16g128)

Quickstart

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "K-Compression/HyperCLOVAX-SEED-Think-14B-GPTQ"
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="bfloat16",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

Performance(Non-Think)

Model	MMLU (0-shot)	HAERAE (0-shot)
HyperCLOVA X SEED 14B Think	0.7144	0.8130
HyperCLOVA X SEED 14B Think-GPTQ	0.7018	0.8139

License

The model is licensed under HyperCLOVA X SEED Model License Agreement