Lingo-IITGN commited on
Commit
8fc0b91
1 Parent(s): 6dd3352

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +36 -14
README.md CHANGED
@@ -81,16 +81,38 @@ This model described is a research preview and is under ongoing iterative updati
81
 
82
  ### Results
83
 
84
-
85
- | Model | Fertility | PPL |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86
  |:-----------:|:---------:|:------:|
87
- | ganga-1b | 1.12 | 34.85 |
88
- | pragna-1b | 1.58 | 12.74 |
89
- | bloom-1b1 | 1.27 | 33.39 |
90
- | bloom-1b7 | 1.27 | 26.63 |
91
- | gemma-2b | 1.89 | 41.67 |
92
- | bloom-3b | 1.27 | 23.77 |
93
- | airavata-7b | 1.69 | 46.24 |
 
 
94
 
95
 
96
  #### Summary
@@ -104,15 +126,15 @@ This model described is a research preview and is under ongoing iterative updati
104
  Ganga-1b is a decoder-only transformer model, featuring the following specifications:
105
 
106
 
107
- * #Layers: 16
108
- * #Attention heads: 32
109
- * Embedding dimension: 2048
110
- * Vocabulary size: 30000
111
  * Sliding window: 512
112
  * Intermediate dimension: 716
113
 
114
 
115
  ## Model Card Contact
116
 
117
- [Lingo Research Labs at IIT Gandhinagar, India](https://labs.iitgn.ac.in/lingo/)
118
 
81
 
82
  ### Results
83
 
84
+ <details open>
85
+ <summary>Tokenizers Results</summary>
86
+ <br>
87
+
88
+ | Model | Fertility |
89
+ |:-----------:|:---------:|
90
+ | ***ganga-1b*** | ***1.12*** |
91
+ | pragna-1b | 1.58 |
92
+ | bloom-1b1 | 1.27 |
93
+ | bloom-1b7 | 1.27 |
94
+ | gemma-2b | 1.89 |
95
+ | bloom-3b | 1.27 |
96
+ | airavata-7b | 1.69 |
97
+
98
+ </details>
99
+
100
+
101
+ <details open>
102
+ <summary>Metrics</summary>
103
+ <br>
104
+
105
+ | Model | PPL_{Ours | PPL_{Airawat} |
106
  |:-----------:|:---------:|:------:|
107
+ | ganga-1b | | 34.85 |
108
+ | pragna-1b | | 12.74 |
109
+ | bloom-1b1 | | 33.39 |
110
+ | bloom-1b7 | | 26.63 |
111
+ | gemma-2b | | 41.67 |
112
+ | bloom-3b | | 23.77 |
113
+ | airavata-7b | | 46.24 |
114
+
115
+ </details>
116
 
117
 
118
  #### Summary
 
126
  Ganga-1b is a decoder-only transformer model, featuring the following specifications:
127
 
128
 
129
+ * Layers: 16
130
+ * Attention heads: 32
131
+ * Embedding dimension: 2,048
132
+ * Vocabulary size: 30,000
133
  * Sliding window: 512
134
  * Intermediate dimension: 716
135
 
136
 
137
  ## Model Card Contact
138
 
139
+ [Lingo Research Labs at IIT Gandhinagar, India](https://labs.iitgn.ac.in/lingo/) </br>
140