Upload metrics for baseline
Browse files
README.md
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- mindchain/wikitext2
|
| 4 |
+
- yahma/alpaca-cleaned
|
| 5 |
+
metrics:
|
| 6 |
+
- perplexity
|
| 7 |
+
- accuracy
|
| 8 |
+
base_model:
|
| 9 |
+
- TinyLlama/TinyLlama_v1.1
|
| 10 |
+
|
| 11 |
+
|
| 12 |
+
model-index:
|
| 13 |
+
- name: TinyLlama_v1.1_mix_wikitext_alpaca_2bit_BitDistiller_baseline
|
| 14 |
+
results:
|
| 15 |
+
- task:
|
| 16 |
+
type: multiple-choice
|
| 17 |
+
name: QA Benchmarking
|
| 18 |
+
dataset:
|
| 19 |
+
type: allenai/arc
|
| 20 |
+
name: ARC-Challenge
|
| 21 |
+
config: challenge
|
| 22 |
+
split: test
|
| 23 |
+
metrics:
|
| 24 |
+
- type: accuracy
|
| 25 |
+
name: Accuracy
|
| 26 |
+
value: 0.2150170648464164
|
| 27 |
+
- type: accuracy
|
| 28 |
+
name: Normalized Accuracy
|
| 29 |
+
value: 0.24573378839590443
|
| 30 |
+
- task:
|
| 31 |
+
type: multiple-choice
|
| 32 |
+
name: QA Benchmarking
|
| 33 |
+
dataset:
|
| 34 |
+
type: hellaswag
|
| 35 |
+
name: HellaSwag
|
| 36 |
+
split: test
|
| 37 |
+
metrics:
|
| 38 |
+
- type: accuracy
|
| 39 |
+
name: Accuracy
|
| 40 |
+
value: 0.3240390360485959
|
| 41 |
+
- type: accuracy
|
| 42 |
+
name: Normalized Accuracy
|
| 43 |
+
value: 0.37333200557657836
|
| 44 |
+
- task:
|
| 45 |
+
type: multiple-choice
|
| 46 |
+
name: QA Benchmarking
|
| 47 |
+
dataset:
|
| 48 |
+
type: piqa
|
| 49 |
+
name: PIQA
|
| 50 |
+
split: validation
|
| 51 |
+
metrics:
|
| 52 |
+
- type: accuracy
|
| 53 |
+
name: Accuracy
|
| 54 |
+
value: 0.6082698585418934
|
| 55 |
+
- type: accuracy
|
| 56 |
+
name: Normalized Accuracy
|
| 57 |
+
value: 0.6071817192600653
|
| 58 |
+
- task:
|
| 59 |
+
type: multiple-choice
|
| 60 |
+
name: QA Benchmarking
|
| 61 |
+
dataset:
|
| 62 |
+
type: winogrande
|
| 63 |
+
name: Winogrande
|
| 64 |
+
split: test
|
| 65 |
+
metrics:
|
| 66 |
+
- type: accuracy
|
| 67 |
+
name: Accuracy
|
| 68 |
+
value: 0.5201262825572218
|
| 69 |
+
- task:
|
| 70 |
+
type: multiple-choice
|
| 71 |
+
name: QA Benchmarking
|
| 72 |
+
dataset:
|
| 73 |
+
type: aggregated
|
| 74 |
+
name: QA-Avg
|
| 75 |
+
metrics:
|
| 76 |
+
- type: accuracy
|
| 77 |
+
name: QA Average
|
| 78 |
+
value: 0.4168630604985319
|
| 79 |
+
- task:
|
| 80 |
+
type: language-modeling
|
| 81 |
+
name: Language Modeling
|
| 82 |
+
dataset:
|
| 83 |
+
type: wikitext
|
| 84 |
+
name: WikiText-2
|
| 85 |
+
split: test
|
| 86 |
+
metrics:
|
| 87 |
+
- type: perplexity
|
| 88 |
+
name: Perplexity
|
| 89 |
+
value: 22.655162811279297
|
| 90 |
+
|
| 91 |
+
|
| 92 |
+
---
|
| 93 |
+
|
| 94 |
+
TODO: check the splits of each dataset
|