Downloads · 30 days
4
9% of all-time downloads
mdeberta_EW-TT-PE_U0_S1_Tingredient_P0.5_DROP1_mdeberta_E4_DEV97.0
pgajo/mdeberta_EW-TT-PE_U0_S1_Tingredient_P0.5_DROP1_mdeberta_E4_DEV97.0
mdeberta_EW-TT-PE_U0_S1_Tingredient_P0.5_DROP1_mdeberta_E4_DEV97.0 is a question answering model from pgajo. Use it when the input is a question plus a passage. It is set up for transformers.
All-time downloads
46
Public
Parameters
278M
1.1 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors1.1 GB · 98%
At a glance
- Task
- Question Answering
- Library
- transformers
- Model type
- deberta-v2
- Access
- Public
- Created
- Feb 13, 2024
- Updated
- Feb 13, 2024
- SHA
- 94a7df7e
Try a prompt
Model card
From the Hugging Face model README
Model description:
Model: microsoft/mdeberta-v3-base
Dataset: TASTEset
Unshuffled ratio: ['0']
Shuffled ratio: ['1']
Best exact match epoch: 4
Best exact match: 96.7
Best epoch: 4
Drop duplicates: ['1']
Max epochs = 10
Optimizer lr = 3e-05
Optimizer eps = 1e-08
Batch size = 8
Dataset path = pgajo/EW-TT-PE_U0_S1_Tingredient_P0.5_DROP1_mdeberta
Results
| epoch | train_loss | train_f1 | train_exact | dev_loss | dev_f1 | dev_exact | test_loss | test_f1 | test_exact |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 1.61 | 59.81 | 52.89 | 0.18 | 96.25 | 93.68 | 0 | 0 | 0 |
| 2 | 0.19 | 96.12 | 94.15 | 0.15 | 96.91 | 95.6 | 0 | 0 | 0 |
| 3 | 0.09 | 97.95 | 97.25 | 0.15 | 96.58 | 95.33 | 0 | 0 | 0 |
| 4 | 0.05 | 99.07 | 98.48 | 0.16 | 97.7 | 96.7 | 0 | 0 | 0 |
| 5 | 0.03 | 99.02 | 98.42 | 0.15 | 97.24 | 95.6 | 0 | 0 | 0 |
| 6 | 0.02 | 99.54 | 99.1 | 0.17 | 97.18 | 95.6 | 0 | 0 | 0 |
| 7 | 0.06 | 98.33 | 97.8 | 0.18 | 96.13 | 94.51 | 0 | 0 | 0 |
- Task
- Question Answering
- Library
- transformers
- Type
- deberta-v2
- Created
- Feb 13, 2024
- Updated
- Feb 13, 2024