Update README.md (#2)
Browse files- Update README.md (6db5b387dc0536e8ffdb6e805971de02049dd79c)
README.md
CHANGED
@@ -20,11 +20,19 @@ pipeline_tag: text-generation
|
|
20 |
- https://huggingface.co/datasets/jakartaresearch/indoqa
|
21 |
|
22 |
|
23 |
-
**Task**:
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
**Experiment**:
|
28 |
-
- Use
|
29 |
-
-
|
30 |
-
-
|
|
|
|
|
|
|
|
20 |
- https://huggingface.co/datasets/jakartaresearch/indoqa
|
21 |
|
22 |
|
23 |
+
**Task**:
|
24 |
+
Chat or Conversational
|
25 |
+
|
26 |
+
**Input**:
|
27 |
+
User's prompt containing chat templated text in string format
|
28 |
+
|
29 |
+
**Output**:
|
30 |
+
Model's generated text in string format
|
31 |
|
32 |
**Experiment**:
|
33 |
+
- Use bos_token and eos_token to replace <|im_start|> and <|im_end|> in ChatML. (Inspired by: https://asmirnov.xyz/doppelganger)
|
34 |
+
- Use left padding and left truncation to conform to max_length.
|
35 |
+
- Set max_length = 256 in the training process, which consumes 33.7 GB of memory.
|
36 |
+
|
37 |
+
**Notebook**:
|
38 |
+
- https://drive.google.com/file/d/11FiaWxGt2HxUirZrHTNLaVmiqrUwejwV/view?usp=drive_link
|