patricia-rocha commited on
Commit
844d844
1 Parent(s): d35f522

Update README.md and add logo

Browse files
Files changed (2) hide show
  1. README.md +17 -6
  2. logo.png +0 -0
README.md CHANGED
@@ -4,11 +4,22 @@ license: cc-by-nc-4.0
4
  co2_eq_emissions: 710
5
  ---
6
 
7
- # QUOKKA
 
 
 
 
 
 
 
 
 
 
 
8
 
9
  ## Model description
10
 
11
- QUOKKA is our first generative pre-trained transformer (GPT) model for Portuguese from Portugal (PT-PT).
12
  Our model is a fine-tuned version of [Phoenix](https://huggingface.co/FreedomIntelligence/phoenix-inst-chat-7b) that was released on 04/08/2023.
13
  The backbone of Phoenix is [BLOOMZ](https://huggingface.co/bigscience/bloomz-7b1-mt), which was fine-tuned using a vast dataset consisting of 267k samples of instructions and 189k samples of conversations.
14
 
@@ -85,7 +96,7 @@ generator(f"<human>{prompt}<bot>", max_new_tokens=512, temperature=0.0)
85
 
86
  ## Training data
87
 
88
- QUOKKA was fine-tuned on a dataset collected from different sources:
89
 
90
  * Initially, we used the **[Cabrita](https://github.com/22-hours/cabrita)** dataset that consists of a translation of Alpaca's training data.
91
  The Portuguese translation was generated using ChatGPT. Therefore, it is important to note that these translations may not be of the highest quality.
@@ -119,9 +130,9 @@ Follows the results against GPT-3.5 and two of the highest performing open-sourc
119
 
120
  | | **Lose** | **Tie** | **Win** |
121
  |------------------------|----------|---------|---------|
122
- | QUOKKA vs. **GPT-3.5** | 63.8% | 10.1% | 26.1% |
123
- | QUOKKA vs. **Vicuna** | 66.2% | 8.8% | 25.0% |
124
- | QUOKKA vs. **Falcon** | 17.4% | 1.4% | 81.2% |
125
 
126
  ## Environmental impact
127
 
 
4
  co2_eq_emissions: 710
5
  ---
6
 
7
+ <img src="logo.png" alt="logo" style="width: 500px" />
8
+
9
+ # Table of Contents
10
+
11
+ 1. [Model description](#model-description)
12
+ 2. [Intended uses & limitations](#intended-uses--limitations)
13
+ 3. [Training data](#training-data)
14
+ 4. [Training procedure](#training-procedure)
15
+ 5. [Evaluation results](#evaluation-results)
16
+ 6. [Environmental impact](#environmental-impact)
17
+
18
+ # Quokka
19
 
20
  ## Model description
21
 
22
+ Quokka is our first generative pre-trained transformer (GPT) model for Portuguese from Portugal (PT-PT).
23
  Our model is a fine-tuned version of [Phoenix](https://huggingface.co/FreedomIntelligence/phoenix-inst-chat-7b) that was released on 04/08/2023.
24
  The backbone of Phoenix is [BLOOMZ](https://huggingface.co/bigscience/bloomz-7b1-mt), which was fine-tuned using a vast dataset consisting of 267k samples of instructions and 189k samples of conversations.
25
 
 
96
 
97
  ## Training data
98
 
99
+ Quokka was fine-tuned on a dataset collected from different sources:
100
 
101
  * Initially, we used the **[Cabrita](https://github.com/22-hours/cabrita)** dataset that consists of a translation of Alpaca's training data.
102
  The Portuguese translation was generated using ChatGPT. Therefore, it is important to note that these translations may not be of the highest quality.
 
130
 
131
  | | **Lose** | **Tie** | **Win** |
132
  |------------------------|----------|---------|---------|
133
+ | Quokka vs. **GPT-3.5** | 63.8% | 10.1% | 26.1% |
134
+ | Quokka vs. **Vicuna** | 66.2% | 8.8% | 25.0% |
135
+ | Quokka vs. **Falcon** | 17.4% | 1.4% | 81.2% |
136
 
137
  ## Environmental impact
138
 
logo.png ADDED