Compare commits

...

10 Commits

Author SHA1 Message Date
Niels Rogge 81916d20ee Upload TF model 2021-11-29 11:41:09 +00:00
kamalkraj bcb81193f6 TF model weights 2021-10-11 13:03:09 +00:00
Niels Rogge 216b8946b0 Update README 2020-12-17 14:59:50 +00:00
Niels Rogge 6a0f63d7d2 Make results header smaller 2020-12-17 14:32:39 +00:00
Niels Rogge 5d8ba8b8f1 Make results header smaller 2020-12-17 14:18:30 +00:00
Niels Rogge 71b4517344 Fix another typo 2020-12-17 14:15:00 +00:00
Niels Rogge eddb779b2d Fix typo 2020-12-17 14:10:38 +00:00
Niels Rogge d376c0e232 Update results table 2020-12-17 14:09:41 +00:00
Niels Rogge a0c6ebb891 Add results table 2020-12-17 14:03:51 +00:00
Niels Rogge 2e5da5cadd Fix name of SQA dataset 2020-12-15 17:40:51 +00:00
3 changed files with 85 additions and 61 deletions

View File

@ -2,10 +2,10 @@
language: en
tags:
- tapas
- question-answering
- table-question-answering
license: apache-2.0
datasets:
- sqa
- msr_sqa
---
# TAPAS base model fine-tuned on Sequential Question Answering (SQA)
@ -19,6 +19,23 @@ The other (non-default) version which can be used is:
Disclaimer: The team releasing TAPAS did not write a model card for this model so this model card has been written by
the Hugging Face team and contributors.
## Results on SQA - Dev Accuracy
Size | Reset | Dev Accuracy | Link
-------- | --------| -------- | ----
LARGE | noreset | 0.7223 | [tapas-large-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-large-finetuned-sqa/tree/no_reset)
LARGE | reset | 0.7289 | [tapas-large-finetuned-sqa](https://huggingface.co/google/tapas-large-finetuned-sqa/tree/main)
**BASE** | **noreset** | **0.6737** | [tapas-base-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-base-finetuned-sqa/tree/no_reset)
**BASE** | **reset** | **0.6874** | [tapas-base-finetuned-sqa](https://huggingface.co/google/tapas-base-finetuned-sqa/tree/main)
MEDIUM | noreset | 0.6464 | [tapas-medium-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-medium-finetuned-sqa/tree/no_reset)
MEDIUM | reset | 0.6561 | [tapas-medium-finetuned-sqa](https://huggingface.co/google/tapas-medium-finetuned-sqa/tree/main)
SMALL | noreset | 0.5876 | [tapas-small-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-small-finetuned-sqa/tree/no_reset)
SMALL | reset | 0.6155 | [tapas-small-finetuned-sqa](https://huggingface.co/google/tapas-small-finetuned-sqa/tree/main)
MINI | noreset | 0.4574 | [tapas-mini-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-mini-finetuned-sqa/tree/no_reset)
MINI | reset | 0.5148 | [tapas-mini-finetuned-sqa](https://huggingface.co/google/tapas-mini-finetuned-sqa/tree/main))
TINY | noreset | 0.2004 | [tapas-tiny-finetuned-sqa (absolute pos embeddings)](https://huggingface.co/google/tapas-tiny-finetuned-sqa/tree/no_reset)
TINY | reset | 0.2375 | [tapas-tiny-finetuned-sqa](https://huggingface.co/google/tapas-tiny-finetuned-sqa/tree/main)
## Model description
TAPAS is a BERT-like transformers model pretrained on a large corpus of English data from Wikipedia in a self-supervised fashion.

View File

@ -1,59 +1,63 @@
{
"aggregation_loss_weight": 1.0,
"aggregation_temperature": 1.0,
"allow_empty_column_selection": false,
"answer_loss_cutoff": null,
"answer_loss_importance": 1.0,
"architectures": [
"TapasForQuestionAnswering"
],
"attention_probs_dropout_prob": 0.1,
"average_approximation_function": "ratio",
"average_logits_per_cell": false,
"cell_selection_preference": null,
"disable_per_token_loss": false,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"huber_loss_delta": null,
"init_cell_selection_weights_to_zero": false,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_num_columns": 32,
"max_num_rows": 64,
"max_position_embeddings": 1024,
"model_type": "tapas",
"num_aggregation_labels": 0,
"num_attention_heads": 12,
"num_hidden_layers": 12,
"pad_token_id": 0,
"positive_label_weight": 10.0,
"reset_position_index_per_cell": true,
"select_one_column": true,
"softmax_temperature": 1.0,
"type_vocab_size": [
3,
256,
256,
2,
256,
256,
10
],
"type_vocab_sizes": [
3,
256,
256,
2,
256,
256,
10
],
"use_answer_as_supervision": null,
"use_gumbel_for_aggregation": false,
"use_gumbel_for_cells": false,
"use_normalized_answer_loss": false,
"vocab_size": 30522
}
{
"_name_or_path": "google/tapas-base-finetuned-sqa",
"aggregation_labels": null,
"aggregation_loss_weight": 1.0,
"aggregation_temperature": 1.0,
"allow_empty_column_selection": false,
"answer_loss_cutoff": null,
"answer_loss_importance": 1.0,
"architectures": [
"TapasForQuestionAnswering"
],
"attention_probs_dropout_prob": 0.1,
"average_approximation_function": "ratio",
"average_logits_per_cell": false,
"cell_selection_preference": null,
"disable_per_token_loss": false,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"huber_loss_delta": null,
"init_cell_selection_weights_to_zero": false,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-12,
"max_num_columns": 32,
"max_num_rows": 64,
"max_position_embeddings": 1024,
"model_type": "tapas",
"no_aggregation_label_index": null,
"num_aggregation_labels": 0,
"num_attention_heads": 12,
"num_hidden_layers": 12,
"pad_token_id": 0,
"positive_label_weight": 10.0,
"reset_position_index_per_cell": true,
"select_one_column": true,
"softmax_temperature": 1.0,
"transformers_version": "4.13.0.dev0",
"type_vocab_size": [
3,
256,
256,
2,
256,
256,
10
],
"type_vocab_sizes": [
3,
256,
256,
2,
256,
256,
10
],
"use_answer_as_supervision": null,
"use_gumbel_for_aggregation": false,
"use_gumbel_for_cells": false,
"use_normalized_answer_loss": false,
"vocab_size": 30522
}

BIN
tf_model.h5 (Stored with Git LFS) Normal file

Binary file not shown.