List your company

Language Modeling on FW (val)

5.84PPLDense
-191.72241,141.82382,475.373,808.9162May 18, 2025

Evaluation results

MethodLinks
DenseSparsity=0%, Latency=-, Model=Llama-2-13B2025.055.84
EvoPressSparsity=12.5%, Latency=29m, Model=Llama-2-13B2025.056.62
UniCuCoSparsity=12.5%, Latency=<1s, Model=Llama-2-13B2025.056.94
Cosine (Window)Sparsity=12.5%, Latency=16s, Model=Llama-2-13B2025.056.97
ShortGPTSparsity=12.5%, Latency=<1s, Model=Llama-2-13B2025.057.19
Weight SubcloningSparsity=12.5%, Latency=<1s, Model=Llama-2-13B2025.057.26
EvoPressSparsity=25%, Latency=28.6m, Model=Llama-2-13B2025.058.32
UniCuCoSparsity=25%, Latency=<1s, Model=Llama-2-13B2025.0510.14
Cosine (Window)Sparsity=25%, Latency=24s, Model=Llama-2-13B2025.0510.34
EvoPressSparsity=37.5%, Latency=26.8m, Model=Llama-2-13B2025.0513
ShortGPTSparsity=25%, Latency=<1s, Model=Llama-2-13B2025.0515.73
Weight SubcloningSparsity=25%, Latency=<1s, Model=Llama-2-13B2025.0515.73
UniCuCoSparsity=37.5%, Latency=<1s, Model=Llama-2-13B2025.0517.74
EvoPressSparsity=50%, Latency=28.5m, Model=Llama-2-13B2025.0528.48
ShortGPTSparsity=37.5%, Latency=<1s, Model=Llama-2-13B2025.0535.24
Weight SubcloningSparsity=37.5%, Latency=<1s, Model=Llama-2-13B2025.0535.24
Cosine (Window)Sparsity=37.5%, Latency=24s, Model=Llama-2-13B2025.0550.46
UniCuCoSparsity=50%, Latency=<1s, Model=Llama-2-13B2025.0576.79
ShortGPTSparsity=50%, Latency=<1s, Model=Llama-2-13B2025.05132.47
Weight SubcloningSparsity=50%, Latency=<1s, Model=Llama-2-13B2025.05132.47
Cosine (Window)Sparsity=50%, Latency=17s, Model=Llama-2-13B2025.05316.09
UniCuCoSparsity=62.5%, Latency=<1s, Model=Llama-2-13B2025.05383.87
ShortGPTSparsity=62.5%, Latency=<1s, Model=Llama-2-13B2025.05622.41
Weight SubcloningSparsity=62.5%, Latency=<1s, Model=Llama-2-13B2025.05622.41
EvoPressSparsity=62.5%, Latency=23.5m, Model=Llama-2-13B2025.053,930.84
Cosine (Window)Sparsity=62.5%, Latency=13s, Model=Llama-2-13B2025.054,944.9