jason9693 commited on
Commit
515df60
β€’
1 Parent(s): 03fc19a

update metric score

Browse files
Files changed (1) hide show
  1. README.md +4 -15
README.md CHANGED
@@ -72,11 +72,11 @@ As with all language models, it is hard to predict in advance how GPT-NeoX-Ko wi
72
 
73
  <figure>
74
 
75
- | Model | Public | Training FLOPs | LAMBADA PPL ↓ | LAMBADA Acc ↑ | Winogrande ↑ | Hellaswag ↑ | PIQA ↑ | Dataset Size (GB) |
76
  |--------------------------|-------------|----------------|--- |--- |--- |--- |--- |-------------------|
77
- | KoGPT-trinity&ddagger; | &cross; | ----- | 3.0 | 75% | 72% | 78% | 80% | ----- |
78
- | KoGPT-KakaoBrain&ddagger; | &cross; | ----- | 3.0 | 75% | 72% | 78% | 80% | ----- |
79
- | GPT-NeoX-Ko-1.3B(ours)&ddagger; | &cross; | ----- | 3.0 | 75% | 72% | 78% | 80% | ----- |
80
 
81
 
82
  <figcaption><p>Models roughly sorted by performance, or by FLOPs if not available.</p>
@@ -111,17 +111,6 @@ To cite this model:
111
  }
112
  ```
113
 
114
- To cite the codebase that trained this model:
115
- ```bibtex
116
- @misc{mesh-transformer-jax,
117
- author = {Wang, Ben},
118
- title = {{Mesh-Transformer-JAX: Model-Parallel Implementation of Transformer Language Model with JAX}},
119
- howpublished = {\url{https://github.com/kingoflolz/mesh-transformer-jax}},
120
- year = 2021,
121
- month = May
122
- }
123
- ```
124
-
125
  If you use this model, we would love to hear about it! Reach out on [GitHub](https://github.com/kingoflolz/mesh-transformer-jax), Discord, or shoot Ben an email.
126
 
127
  ## Acknowledgements
 
72
 
73
  <figure>
74
 
75
+ | Model | Public | Training FLOPs | kobest_boolq ↓ | kobest_copa ↑ | kobest_wic ↑ | kobest_hellaswag ↑ | kobest_sentineg ↑ | Dataset Size (GB) |
76
  |--------------------------|-------------|----------------|--- |--- |--- |--- |--- |-------------------|
77
+ | KoGPT-trinity&ddagger; | &cross; | ----- | 0.6663 | 0.6222 | 0.656 | 0.4011 | 0.3534 | ----- |
78
+ | KoGPT-KakaoBrain&ddagger; | &cross; | ----- | 0.3241 | 0.719 | 0.1356 | 0.4616 | 0.8065 | ----- |
79
+ | GPT-NeoX-Ko-1.3B(ours)&ddagger; | &cross; | ----- | 0.5174 | 0.7072 | 0.6567 | 0.417 | 0.8444 | ----- |
80
 
81
 
82
  <figcaption><p>Models roughly sorted by performance, or by FLOPs if not available.</p>
 
111
  }
112
  ```
113
 
 
 
 
 
 
 
 
 
 
 
 
114
  If you use this model, we would love to hear about it! Reach out on [GitHub](https://github.com/kingoflolz/mesh-transformer-jax), Discord, or shoot Ben an email.
115
 
116
  ## Acknowledgements