Update README.md
Browse files
README.md
CHANGED
@@ -83,3 +83,16 @@ configs:
|
|
83 |
- split: benchmark_results
|
84 |
path: data/train-*
|
85 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
83 |
- split: benchmark_results
|
84 |
path: data/train-*
|
85 |
---
|
86 |
+
|
87 |
+
# Analysis of energy usage for HUGS models
|
88 |
+
|
89 |
+
Based on the [energy_star branch](https://github.com/huggingface/optimum-benchmark/tree/energy_star_dev) of [optimum-benchmark](https://github.com/huggingface/optimum-benchmark), and using [codecarbon](https://pypi.org/project/codecarbon/2.1.4/).
|
90 |
+
|
91 |
+
# Fields
|
92 |
+
- task: Task the model was benchmarked on
|
93 |
+
- org: Organization hosting the model
|
94 |
+
- model: The specific model. Model names at HF are usually constructed with {org}/{model}
|
95 |
+
- date: The date that the benchmark was fun
|
96 |
+
- prefill: The esimated energy and efficiency for prefilling.
|
97 |
+
- decode: The estimated energy and efficiency for decoding.
|
98 |
+
- preprocess: The estimated energy and efficiency for preprocessing.
|