init.

README.md

@@ -0,0 +1,53 @@
+---
+tags:
+- question-answering, multi-step-reasoning, multi-hop-reasoning
+thumbnail: https://raw.githubusercontent.com/StonyBrookNLP/teabreac/main/teabreac_icon.png
+license: cc-by-4.0
+---
+
+# What's this?
+
+This is one of the models reported in the paper: ["Teaching Broad Reasoning Skills for Multi-Step QA by Generating Hard Contexts".](https://arxiv.org/abs/2205.12496).
+
+This paper proposes a procedure to synthetically generate a QA dataset, TeaBReaC, for pretraining language models for robust multi-step reasoning. Pretraining plain LMs like Bart, T5 and numerate LMs like NT5, PReasM, POET on TeaBReaC leads to improvemed downstream performance on several multi-step QA datasets. Please checkout out the paper for the details.
+
+We release the following models:
+
+- **A:** Base Models finetuned on target datasets: `{target_dataset}-{base_model}`
+- **B:** Base models pretrained on TeaBReaC: `teabreac-{base_model}`
+- **C:** Base models pretrained on TeaBReaC and then finetuned on target datasets: `teabreac-{target_dataset}-{base_model}`
+
+The `base_model` above can be from: `bart-large`, `t5-large`, `t5-3b`, `nt5-small`, `preasm-large`.
+The `target_dataset` above can be from: `drop`, `tatqa`, `iirc-gold`, `iirc-retrieved`, `numglue`.
+
+The **A** models are only released for completeness / reproducibility. In your end application you probably just want to use either **B** or **C**.
+
+# How to use it?
+
+Please checkout the details in our [github repository](https://github.com/stonybrooknlp/teabreac), but in a nutshell:
+
+```python
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+from digit_tokenization import enable_digit_tokenization # digit_tokenization.py from https://github.com/stonybrooknlp/teabreac
+
+model_name = "poet-large-tatqa"
+tokenizer = AutoTokenizer.from_pretrained(model_name, use_fast=False) # Fast doesn't work with digit tokenization
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+enable_digit_tokenization(tokenizer)
+input_texts = [
+    "answer_me: Who scored the first touchdown of the game?" +
+    "context: ... Oakland would get the early lead in the first quarter as quarterback JaMarcus Russell completed a 20-yard touchdown pass to rookie wide receiver Chaz Schilens..."
+    # Note: some models have slightly different qn/ctxt format. See the github repo.
+]
+input_ids = tokenizer(
+    input_texts, return_tensors="pt",
+    truncation=True, max_length=800,
+    add_special_tokens=True, padding=True,
+)
+generated_ids = model.generate(input_ids, min_length=1,  max_length=50)
+generated_predictions = tokenizer.batch_decode(generated_ids, skip_special_tokens=False)
+generated_predictions = [
+    tokenizer.fix_decoded_text(generated_prediction) for generated_prediction in generated_predictions
+]
+# => ["Chaz Schilens"]
+```

added_tokens.json

@@ -0,0 +1,66 @@
+{
+  "<ss>": 50309,
+  "[CHPD]": 50313,
+  "[CND]": 50310,
+  "[COLD]": 50303,
+  "[CTANS]": 50314,
+  "[CTCNINSINA]": 50318,
+  "[CTCNINTA]": 50317,
+  "[CTOUINSINA]": 50316,
+  "[CTOUINTA]": 50315,
+  "[DAY]": 50326,
+  "[DFB]": 50305,
+  "[DFD]": 50307,
+  "[DFE]": 50306,
+  "[MDFD]": 50308,
+  "[MONTH]": 50327,
+  "[ROWD]": 50304,
+  "[SCBN]": 50323,
+  "[SCBP]": 50322,
+  "[SCD]": 50321,
+  "[SCMN]": 50325,
+  "[SCMP]": 50324,
+  "[SIND]": 50312,
+  "[SNE]": 50320,
+  "[SNS]": 50319,
+  "[SOMD]": 50311,
+  "[YEAR]": 50328,
+  "are_items_different": 50267,
+  "are_items_same": 50266,
+  "arg_bool": 50295,
+  "arg_intersection": 50296,
+  "arg_maximum_date": 50280,
+  "arg_maximum_number": 50269,
+  "arg_minimum_date": 50279,
+  "arg_minimum_number": 50268,
+  "arithmetic_division": 50275,
+  "arithmetic_mean_single": 50272,
+  "arithmetic_multiplication": 50277,
+  "arithmetic_subtraction": 50276,
+  "arithmetic_sum_multiple": 50274,
+  "arithmetic_sum_single": 50273,
+  "boolean": 50265,
+  "compare_dates": 50284,
+  "compare_numbers": 50278,
+  "date_subtraction": 50283,
+  "filter_a_where_b_is_compared_to": 50292,
+  "filter_a_where_b_is_compared_to_date": 50288,
+  "filter_a_where_b_is_given_value": 50291,
+  "filter_a_where_b_is_max": 50289,
+  "filter_a_where_b_is_max_date": 50286,
+  "filter_a_where_b_is_min": 50290,
+  "filter_a_where_b_is_min_date": 50287,
+  "grouped_count": 50298,
+  "grouped_mean": 50299,
+  "grouped_sum": 50300,
+  "intersection": 50297,
+  "kth_highest": 50301,
+  "kth_lowest": 50302,
+  "list_subtraction": 50285,
+  "logical_and": 50294,
+  "logical_or": 50293,
+  "maximum_date": 50282,
+  "maximum_number": 50271,
+  "minimum_date": 50281,
+  "minimum_number": 50270
+}

config.json

@@ -0,0 +1,72 @@
+{
+  "_name_or_path": "SivilTaram/poet-sql-digit",
+  "activation_dropout": 0.1,
+  "activation_function": "gelu",
+  "add_bias_logits": false,
+  "add_final_layer_norm": false,
+  "architectures": [
+    "BartForConditionalGeneration"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 0,
+  "classif_dropout": 0.1,
+  "classifier_dropout": 0.0,
+  "d_model": 1024,
+  "decoder_attention_heads": 16,
+  "decoder_ffn_dim": 4096,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 12,
+  "decoder_start_token_id": 0,
+  "dropout": 0.1,
+  "early_stopping": true,
+  "encoder_attention_heads": 16,
+  "encoder_ffn_dim": 4096,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 12,
+  "eos_token_id": 2,
+  "forced_eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2"
+  },
+  "init_std": 0.02,
+  "is_encoder_decoder": true,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2
+  },
+  "max_position_embeddings": 1024,
+  "model_type": "bart",
+  "normalize_before": false,
+  "num_beams": 4,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "scale_embedding": false,
+  "task_specific_params": {
+    "summarization": {
+      "length_penalty": 1.0,
+      "max_length": 128,
+      "min_length": 12,
+      "num_beams": 4
+    },
+    "summarization_cnn": {
+      "length_penalty": 2.0,
+      "max_length": 142,
+      "min_length": 56,
+      "num_beams": 4
+    },
+    "summarization_xsum": {
+      "length_penalty": 1.0,
+      "max_length": 62,
+      "min_length": 11,
+      "num_beams": 6
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.24.0",
+  "use_cache": true,
+  "vocab_size": 50329
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d135f2011957d9d5714f08fc19b291dc0c9f3e86a921b4a7a864ad6ca278703b
+size 1625793025

special_tokens_map.json

@@ -0,0 +1,63 @@
+{
+  "additional_special_tokens": [
+    "0",
+    "1",
+    "2",
+    "3",
+    "4",
+    "5",
+    "6",
+    "7",
+    "8",
+    "9"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,65 @@
+{
+  "add_prefix_space": false,
+  "add_special_tokens": false,
+  "bos_token": {
+    "__type": "AddedToken",
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "__type": "AddedToken",
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "__type": "AddedToken",
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "errors": "replace",
+  "mask_token": {
+    "__type": "AddedToken",
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "model_max_length": 1024,
+  "name_or_path": "SivilTaram/poet-sql-digit",
+  "pad_token": {
+    "__type": "AddedToken",
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "__type": "AddedToken",
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "special_tokens_map_file": null,
+  "tokenizer_class": "BartTokenizer",
+  "unk_token": {
+    "__type": "AddedToken",
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}