HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks

gemma-2-27b

google/gemma-2-27b

kind
original
quant
bf16 (16-bit)
family
gemma
modality
text-generation
root model
google/gemma-2-27b
provider
google
license
gemma
language
—
context
,
library
transformers
vRAM (GB)
—
size on disk
—
downloads
5,933
likes
209
created
—
last modified
2024-08-07T18:26:09+00:00 (2024)
huggingface
https://huggingface.co/google/gemma-2-27b
tags
transformerssafetensorsgemma2text-generationarxiv:2009.03300arxiv:1905.07830arxiv:1911.11641arxiv:1904.09728arxiv:1905.10044arxiv:1907.10641arxiv:1811.00937arxiv:1809.02789arxiv:1911.01547arxiv:1705.03551arxiv:2107.03374arxiv:2108.07732arxiv:2110.14168arxiv:2009.11462arxiv:2101.11718arxiv:2110.08193arxiv:1804.09301arxiv:2109.07958arxiv:1804.06876arxiv:2103.03874

Benchmark scores (13)

benchmarkcategoryscorecontextconfevidence
BoolQknowledge84.2[BoolQ][]medium| [BoolQ][boolq] | 0-shot | 84.2 | 84.8 |
HellaSwagknowledge81.9[HellaSwag][]medium| [HellaSwag][hellaswag] | 10-shot | 81.9 | 86.4 |
PIQAknowledge81.7[PIQA][]medium| [PIQA][piqa] | 0-shot | 81.7 | 83.2 |
WinoGrandeknowledge80.6[WinoGrande][]medium| [WinoGrande][winogrande] | partial score | 80.6 | 83.7 |
TriviaQAknowledge76.6[TriviaQA][]medium| [TriviaQA][triviaqa] | 5-shot | 76.6 | 83.7 |
MMLUknowledge71.3[MMLU][]medium| [MMLU][mmlu] | 5-shot, top-1 | 71.3 | 75.2 |
GSM8Kmath68.6[GSM8K][]medium| [GSM8K][gsm8k] | 5-shot, maj@1 | 68.6 | 74.0 |
BigBenchreasoning68.2[BIG-Bench][big-bench]medium| [BIG-Bench][big-bench] | 3-shot, CoT | 68.2 | 74.9 |
AGIEvalreasoning52.8[AGIEval][]medium| [AGIEval][agieval] | 3-5-shot | 52.8 | 55.1 |
MBPPcoding52.4[MBPP][]medium| [MBPP][mbpp] | 3-shot | 52.4 | 62.6 |
TruthfulQAsafety50.27[TruthfulQA][]medium| [TruthfulQA][truthfulqa] | | 50.27 | 51.60 |
HumanEvalcoding40.2[HumanEval][]medium| [HumanEval][humaneval] | pass@1 | 40.2 | 51.8 |
MATHmath36.6[MATH][]medium| [MATH][math] | 4-shot | 36.6 | 42.3 |

Discussions (0)

none scraped.

Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.