Update README.md
Browse files
README.md
CHANGED
@@ -3,57 +3,71 @@ license: other
|
|
3 |
base_model: facebook/opt-350m
|
4 |
tags:
|
5 |
- generated_from_trainer
|
|
|
|
|
|
|
6 |
metrics:
|
7 |
- accuracy
|
8 |
model-index:
|
9 |
- name: mini3
|
10 |
results: []
|
|
|
|
|
|
|
11 |
---
|
12 |
|
13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
14 |
should probably proofread and complete it, then remove this comment. -->
|
15 |
|
16 |
-
#
|
17 |
|
18 |
-
|
19 |
-
It achieves the following results on the evaluation set:
|
20 |
-
- Loss: 4.4915
|
21 |
-
- Accuracy: 0.3897
|
22 |
|
23 |
-
|
24 |
|
25 |
-
|
26 |
|
27 |
-
|
28 |
|
29 |
-
|
|
|
30 |
|
31 |
-
## Training and evaluation data
|
32 |
|
33 |
-
|
34 |
|
35 |
-
|
|
|
36 |
|
37 |
-
### Training hyperparameters
|
38 |
|
39 |
-
The
|
40 |
-
- learning_rate: 0.0001
|
41 |
-
- train_batch_size: 1
|
42 |
-
- eval_batch_size: 8
|
43 |
-
- seed: 42
|
44 |
-
- gradient_accumulation_steps: 32
|
45 |
-
- total_train_batch_size: 32
|
46 |
-
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
47 |
-
- lr_scheduler_type: constant
|
48 |
-
- num_epochs: 75.0
|
49 |
|
50 |
-
|
51 |
|
|
|
|
|
52 |
|
|
|
|
|
53 |
|
54 |
-
###
|
|
|
55 |
|
56 |
-
|
57 |
-
|
58 |
-
-
|
59 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
base_model: facebook/opt-350m
|
4 |
tags:
|
5 |
- generated_from_trainer
|
6 |
+
- qa
|
7 |
+
- open data
|
8 |
+
- opt
|
9 |
metrics:
|
10 |
- accuracy
|
11 |
model-index:
|
12 |
- name: mini3
|
13 |
results: []
|
14 |
+
datasets:
|
15 |
+
- ccore/open_data_understanding
|
16 |
+
pipeline_tag: text-generation
|
17 |
---
|
18 |
|
19 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
20 |
should probably proofread and complete it, then remove this comment. -->
|
21 |
|
22 |
+
# OPT_350_open_data_understanding
|
23 |
|
24 |
+
## Description
|
|
|
|
|
|
|
25 |
|
26 |
+
This model has been trained to understand and respond to any content inserted after the `[PAPER]` tag. It uses advanced language modeling techniques to understand the context, structure, and underlying goals of the input text.
|
27 |
|
28 |
+
## How to use
|
29 |
|
30 |
+
To interact with this template, place your text after the `[PAPER]` tag. The model will process the text and respond accordingly. For example:
|
31 |
|
32 |
+
[PAPER]
|
33 |
+
Your text here...
|
34 |
|
|
|
35 |
|
36 |
+
## Example
|
37 |
|
38 |
+
[PAPER]
|
39 |
+
We present a scalable method to build a high-quality instruction-following language model...
|
40 |
|
|
|
41 |
|
42 |
+
The model will understand and respond to your text according to its context and content.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
+
## Comprehension Sections
|
45 |
|
46 |
+
### [UNDERSTANDING]
|
47 |
+
This section provides a detailed analysis and decomposition of the inserted text, facilitating the understanding of the content.
|
48 |
|
49 |
+
### [QUESTIONS AND ANSWERS]
|
50 |
+
This section addresses questions and answers that could arise based on the text provided.
|
51 |
|
52 |
+
### [OBJECTION AND REPLY]
|
53 |
+
This section addresses any objections and responses that could arise from analysis of the text.
|
54 |
|
55 |
+
## Common questions
|
56 |
+
|
57 |
+
- **What can this model do?**
|
58 |
+
- This model can understand and respond to any text placed after the `[PAPER]` tag.
|
59 |
+
|
60 |
+
- **Is a specific format necessary?**
|
61 |
+
- No, the model is quite flexible regarding the text format.
|
62 |
+
|
63 |
+
- **How does this model perform?**
|
64 |
+
- The model outperforms other LLaMa-based models on the Alpaca leaderboard, demonstrating a highly effective alignment.
|
65 |
+
|
66 |
+
## Warnings
|
67 |
+
|
68 |
+
- This model was trained on a diverse corpus, but may still have bias or limitations.
|
69 |
+
- Continuous validation of the model and its output is essential.
|
70 |
+
|
71 |
+
## Contact and Support
|
72 |
+
|
73 |
+
For more information, visit [Hugging Face](https://huggingface.co/).
|