Upload TF model

README.md

@@ -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. 

config.json

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