List your company

Language Modeling on wikitext (Perplexity and Delta)

8.0668Perplexity (word)Source
7.46435611.53085315.5973519.663847Jul 12, 2024

Evaluation results

MethodLinks
SourceModel=Llama-3.1-8B, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.078.0668—
FlashNormModel=Llama-3.1-8B, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.078.06720.0004
SourceModel=Llama-3.2-1B, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.0712.8837—
FlashNormModel=Llama-3.2-1B, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.0712.88420.0005
SourceModel=SmolLM2-135M, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.0723.1262—
FlashNormModel=SmolLM2-135M, Context window=2048-token, Precision=fp16, Hardware=NVIDIA A100, Evaluation framework=lm-evaluation-harness2024.0723.12790.0017