Update README.md
Browse files
README.md
CHANGED
@@ -69,6 +69,8 @@ widget:
|
|
69 |
|
70 |
# flan-t5-small-instructiongen
|
71 |
|
|
|
|
|
72 |
This model is a fine-tuned version of [google/flan-t5-small](https://huggingface.co/google/flan-t5-small) on the None dataset.
|
73 |
It achieves the following results on the evaluation set:
|
74 |
- Loss: 1.3401
|
@@ -78,17 +80,17 @@ It achieves the following results on the evaluation set:
|
|
78 |
- Rougelsum: 50.338
|
79 |
- Gen Len: 14.0450
|
80 |
|
81 |
-
## Model description
|
82 |
-
|
83 |
-
More information needed
|
84 |
-
|
85 |
## Intended uses & limitations
|
86 |
|
87 |
-
|
|
|
|
|
88 |
|
89 |
## Training and evaluation data
|
90 |
|
91 |
-
|
|
|
|
|
92 |
|
93 |
## Training procedure
|
94 |
|
|
|
69 |
|
70 |
# flan-t5-small-instructiongen
|
71 |
|
72 |
+
Instead of generating questions from text, generate instructions for LLMs!
|
73 |
+
|
74 |
This model is a fine-tuned version of [google/flan-t5-small](https://huggingface.co/google/flan-t5-small) on the None dataset.
|
75 |
It achieves the following results on the evaluation set:
|
76 |
- Loss: 1.3401
|
|
|
80 |
- Rougelsum: 50.338
|
81 |
- Gen Len: 14.0450
|
82 |
|
|
|
|
|
|
|
|
|
83 |
## Intended uses & limitations
|
84 |
|
85 |
+
This is just a **small** model/example. There is likely to be even better performance with larger models (ex [pszemraj/bart-base-instructiongen)](https://huggingface.co/pszemraj/bart-base-instructiongen) generalizes better)
|
86 |
+
|
87 |
+
Additionally, this was trained on a dataset of **only** instructions+outputs, with the `inputs` filtered out. This means that text of *1) cookies and cream 2) chocolate chip 3) mint chip 4) oreo* will **not** get you *"Rank the following ice cream flavors: oreo, mint chip, chocolate chip, cookies and cream"*.
|
88 |
|
89 |
## Training and evaluation data
|
90 |
|
91 |
+
See the linked dataset `pszemraj/fleece2instructions` - it is a filtered/formatted version of `tatsu-lab/alpaca` to generate instructions for arbitrary text.
|
92 |
+
|
93 |
+
- Some of the API examples are intentionally weird to demonstrate the generalizability of the model.
|
94 |
|
95 |
## Training procedure
|
96 |
|