Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -41,7 +41,7 @@ with gr.Blocks() as demo:
|
|
41 |
|
42 |
This repo is the offical demo space for [Scaling Laws with Vocabulary: Larger Models Deserve Larger Vocabularies](https://huggingface.co/papers/2407.13623). In summary, we show that when scaling up model size, increase vocabulary size too, but at a slower rate than other parameters.
|
43 |
|
44 |
-
 as demo:
|
|
56 |
|
57 |
|
58 |
with gr.Row():
|
59 |
-
Nnv = gr.Textbox(label="Non-vocabulary Parameters",
|
60 |
flops = gr.Textbox(label="FLOPs", placeholder="Optional (e.g. 7.05e21)")
|
61 |
output_text = gr.Textbox(label="Prediction")
|
62 |
with gr.Row():
|
|
|
41 |
|
42 |
This repo is the offical demo space for [Scaling Laws with Vocabulary: Larger Models Deserve Larger Vocabularies](https://huggingface.co/papers/2407.13623). In summary, we show that when scaling up model size, increase vocabulary size too, but at a slower rate than other parameters.
|
43 |
|
44 |
+

|
45 |
|
46 |
This tool is used to predict the optimal vocabulary size given the non-vocabulary parameters. We provide 3 ways for prediction:
|
47 |
|
|
|
56 |
|
57 |
|
58 |
with gr.Row():
|
59 |
+
Nnv = gr.Textbox(label="Non-vocabulary Parameters (1e9=1B)", placeholder="Required (e.g. 3.0e9)")
|
60 |
flops = gr.Textbox(label="FLOPs", placeholder="Optional (e.g. 7.05e21)")
|
61 |
output_text = gr.Textbox(label="Prediction")
|
62 |
with gr.Row():
|