List your company

Training Throughput on GPT2-1.5B

4.1ThroughputDirectQ
0.3561.3282.33.272Jun 2, 2022

Evaluation results

MethodLinks
DirectQNetwork Bandwidth=10 Gbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.064.1
DirectQNetwork Bandwidth=10 Gbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.064
DirectQNetwork Bandwidth=1 Gbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.064
DirectQNetwork Bandwidth=1 Gbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.064
AQ-SGDNetwork Bandwidth=10 Gbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.064
AQ-SGDNetwork Bandwidth=10 Gbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.064
AQ-SGDNetwork Bandwidth=1 Gbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.064
DirectQNetwork Bandwidth=500 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.9
DirectQNetwork Bandwidth=500 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063.9
DirectQNetwork Bandwidth=300 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.9
AQ-SGDNetwork Bandwidth=1 Gbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063.9
AQ-SGDNetwork Bandwidth=500 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.9
AQ-SGDNetwork Bandwidth=500 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063.9
FP32Network Bandwidth=10 Gbps, Precision=Full (FP32)2022.063.8
DirectQNetwork Bandwidth=300 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063.8
AQ-SGDNetwork Bandwidth=300 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.8
AQ-SGDNetwork Bandwidth=300 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063.8
DirectQNetwork Bandwidth=100 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.5
AQ-SGDNetwork Bandwidth=100 Mbps, Forward quantization (fw)=3-bit, Backward quantization (bw)=6-bit2022.063.4
FP32Network Bandwidth=1 Gbps, Precision=Full (FP32)2022.063.2
DirectQNetwork Bandwidth=100 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063
AQ-SGDNetwork Bandwidth=100 Mbps, Forward quantization (fw)=4-bit, Backward quantization (bw)=8-bit2022.063
FP32Network Bandwidth=500 Mbps, Precision=Full (FP32)2022.062.7
FP32Network Bandwidth=300 Mbps, Precision=Full (FP32)2022.061.8
FP32Network Bandwidth=100 Mbps, Precision=Full (FP32)2022.060.5