amd
/

Deepseek-S3_sq_a05_v2_mlperf6_1

8-bit precision

Model card Files Files and versions

Quark Team DeepSeek-R1 Model Overview

Model Information For MLPerf

Model Name: deepseek-ai/DeepSeek-R1
Version: MLPerf v6.1
Commit: Close Division Commit

Calibration Dataset

The calibration dataset consists of 500 samples of calibration dataset provided by MLPerf

Quantized Tensors

The following tensors are quantized in each decoder:

MOE Weights
Output projection GEMMs

Ignored Layers

The following layers are ignored during quantization:

lm_head
q_a_proj
q_b_proj
kv_a_proj_with_mqa
kv_b_proj
q_a_proj
q_b_proj

License

Modifications copyright(c) 2024 Advanced Micro Devices,Inc. All rights reserved.

Downloads last month: 22

Safetensors

Model size

339B params

Tensor type

BF16

·

U8

·

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for amd/Deepseek-S3_sq_a05_v2_mlperf6_1

Base model

deepseek-ai/DeepSeek-R1

Quantized

(71)

this model