Update README.md
Browse files
README.md
CHANGED
|
@@ -28,6 +28,87 @@ djinn is a merge of the following models using [LazyMergekit](https://colab.rese
|
|
| 28 |
* paulml/DPOB-INMTOB-7B
|
| 29 |
* mlabonne/AlphaMonarch-7B
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
## 🧩 Configuration
|
| 32 |
Inspired by [theprofessor's config](https://huggingface.co/abacusai/TheProfessor-155b)
|
| 33 |
|
|
|
|
| 28 |
* paulml/DPOB-INMTOB-7B
|
| 29 |
* mlabonne/AlphaMonarch-7B
|
| 30 |
|
| 31 |
+
# 🏆 Benchmarks
|
| 32 |
+
Nous benchmarks, find more [details here](https://gist.github.com/majacinka/3f2a797c8872ca9bfdaa2bbf3369edb5)
|
| 33 |
+
|
| 34 |
+
| Model |AGIEval|GPT4All|TruthfulQA|Bigbench|Average|
|
| 35 |
+
|---------------------------------------------------------------------|------:|------:|---------:|-------:|------:|
|
| 36 |
+
|[chatty-djinn-14B](https://huggingface.co/mayacinka/chatty-djinn-14B)| 38.43| 76.29| 68.02| 47.6| 57.59|
|
| 37 |
+
|
| 38 |
+
### AGIEval
|
| 39 |
+
| Task |Version| Metric |Value| |Stderr|
|
| 40 |
+
|------------------------------|------:|--------|----:|---|-----:|
|
| 41 |
+
|agieval_aqua_rat | 0|acc |23.62|± | 2.67|
|
| 42 |
+
| | |acc_norm|21.65|± | 2.59|
|
| 43 |
+
|agieval_logiqa_en | 0|acc |32.26|± | 1.83|
|
| 44 |
+
| | |acc_norm|33.79|± | 1.86|
|
| 45 |
+
|agieval_lsat_ar | 0|acc |23.04|± | 2.78|
|
| 46 |
+
| | |acc_norm|23.04|± | 2.78|
|
| 47 |
+
|agieval_lsat_lr | 0|acc |38.82|± | 2.16|
|
| 48 |
+
| | |acc_norm|39.22|± | 2.16|
|
| 49 |
+
|agieval_lsat_rc | 0|acc |59.48|± | 3.00|
|
| 50 |
+
| | |acc_norm|54.65|± | 3.04|
|
| 51 |
+
|agieval_sat_en | 0|acc |75.73|± | 2.99|
|
| 52 |
+
| | |acc_norm|74.27|± | 3.05|
|
| 53 |
+
|agieval_sat_en_without_passage| 0|acc |35.92|± | 3.35|
|
| 54 |
+
| | |acc_norm|34.47|± | 3.32|
|
| 55 |
+
|agieval_sat_math | 0|acc |31.36|± | 3.14|
|
| 56 |
+
| | |acc_norm|26.36|± | 2.98|
|
| 57 |
+
|
| 58 |
+
Average: 38.43%
|
| 59 |
+
|
| 60 |
+
### GPT4All
|
| 61 |
+
| Task |Version| Metric |Value| |Stderr|
|
| 62 |
+
|-------------|------:|--------|----:|---|-----:|
|
| 63 |
+
|arc_challenge| 0|acc |62.12|± | 1.42|
|
| 64 |
+
| | |acc_norm|65.44|± | 1.39|
|
| 65 |
+
|arc_easy | 0|acc |83.88|± | 0.75|
|
| 66 |
+
| | |acc_norm|78.58|± | 0.84|
|
| 67 |
+
|boolq | 1|acc |88.07|± | 0.57|
|
| 68 |
+
|hellaswag | 0|acc |65.18|± | 0.48|
|
| 69 |
+
| | |acc_norm|86.45|± | 0.34|
|
| 70 |
+
|openbookqa | 0|acc |39.60|± | 2.19|
|
| 71 |
+
| | |acc_norm|48.60|± | 2.24|
|
| 72 |
+
|piqa | 0|acc |82.26|± | 0.89|
|
| 73 |
+
| | |acc_norm|83.62|± | 0.86|
|
| 74 |
+
|winogrande | 0|acc |83.27|± | 1.05|
|
| 75 |
+
|
| 76 |
+
Average: 76.29%
|
| 77 |
+
|
| 78 |
+
### TruthfulQA
|
| 79 |
+
| Task |Version|Metric|Value| |Stderr|
|
| 80 |
+
|-------------|------:|------|----:|---|-----:|
|
| 81 |
+
|truthfulqa_mc| 1|mc1 |50.55|± | 1.75|
|
| 82 |
+
| | |mc2 |68.02|± | 1.52|
|
| 83 |
+
|
| 84 |
+
Average: 68.02%
|
| 85 |
+
|
| 86 |
+
### Bigbench
|
| 87 |
+
| Task |Version| Metric |Value| |Stderr|
|
| 88 |
+
|------------------------------------------------|------:|---------------------|----:|---|-----:|
|
| 89 |
+
|bigbench_causal_judgement | 0|multiple_choice_grade|57.89|± | 3.59|
|
| 90 |
+
|bigbench_date_understanding | 0|multiple_choice_grade|64.50|± | 2.49|
|
| 91 |
+
|bigbench_disambiguation_qa | 0|multiple_choice_grade|32.56|± | 2.92|
|
| 92 |
+
|bigbench_geometric_shapes | 0|multiple_choice_grade|26.18|± | 2.32|
|
| 93 |
+
| | |exact_str_match | 1.11|± | 0.55|
|
| 94 |
+
|bigbench_logical_deduction_five_objects | 0|multiple_choice_grade|30.80|± | 2.07|
|
| 95 |
+
|bigbench_logical_deduction_seven_objects | 0|multiple_choice_grade|22.86|± | 1.59|
|
| 96 |
+
|bigbench_logical_deduction_three_objects | 0|multiple_choice_grade|57.67|± | 2.86|
|
| 97 |
+
|bigbench_movie_recommendation | 0|multiple_choice_grade|62.00|± | 2.17|
|
| 98 |
+
|bigbench_navigate | 0|multiple_choice_grade|56.20|± | 1.57|
|
| 99 |
+
|bigbench_reasoning_about_colored_objects | 0|multiple_choice_grade|65.65|± | 1.06|
|
| 100 |
+
|bigbench_ruin_names | 0|multiple_choice_grade|64.73|± | 2.26|
|
| 101 |
+
|bigbench_salient_translation_error_detection | 0|multiple_choice_grade|17.33|± | 1.20|
|
| 102 |
+
|bigbench_snarks | 0|multiple_choice_grade|76.24|± | 3.17|
|
| 103 |
+
|bigbench_sports_understanding | 0|multiple_choice_grade|75.15|± | 1.38|
|
| 104 |
+
|bigbench_temporal_sequences | 0|multiple_choice_grade|48.90|± | 1.58|
|
| 105 |
+
|bigbench_tracking_shuffled_objects_five_objects | 0|multiple_choice_grade|22.32|± | 1.18|
|
| 106 |
+
|bigbench_tracking_shuffled_objects_seven_objects| 0|multiple_choice_grade|18.17|± | 0.92|
|
| 107 |
+
|bigbench_tracking_shuffled_objects_three_objects| 0|multiple_choice_grade|57.67|± | 2.86|
|
| 108 |
+
|
| 109 |
+
Average: 47.6%
|
| 110 |
+
|
| 111 |
+
Average score: 57.59%
|
| 112 |
## 🧩 Configuration
|
| 113 |
Inspired by [theprofessor's config](https://huggingface.co/abacusai/TheProfessor-155b)
|
| 114 |
|