Upload 33 files

.gitattributes

@@ -0,0 +1,13 @@
+data/test_data/tests_seen_table_seq_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/test_data/tests_unseen_table_seq_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_1_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_1_tapex.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_1.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_2_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_2_tapex.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train_table_seq_2.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train_data/train.jsonl filter=lfs diff=lfs merge=lfs -text
+data/train.jsonl filter=lfs diff=lfs merge=lfs -text
+data/valid_data/valid_seen_table_seq_iter.jsonl filter=lfs diff=lfs merge=lfs -text
+data/valid_data/valid_unseen_table_seq_iter.jsonl filter=lfs diff=lfs merge=lfs -text

data/convert_to_hf.py

@@ -0,0 +1,32 @@
+import json
+from tabulate import tabulate
+
+
+
+def load_data(filepath):
+    data = []
+    with open(filepath) as f:
+        for line in f.read().splitlines():
+            instance = {}
+            item = json.loads(line)
+            instance["id"] = item["id"]
+            prefix = "SCENE DESCRIPTION:"
+            if prefix not in item["input"]:
+                continue
+            index = item["input"].index(prefix)
+            instance["input_task"] = item["input"][:index].strip()
+            instance["input_table"] = item["input"][index+len(prefix):].strip().split("[SEP]")
+            instance["input_table"] = [r.strip().split(",") for r in instance["input_table"]]
+            instance["input_table_md"] = tabulate(instance["input_table"],headers='firstrow')
+            instance["output_text"] = item["output"]
+            instance["output_list"] = [o.split("|") for o in item["output"]]
+            instance["label"] = "unseen" if "unseen" in filepath else "seen"
+            data.append(instance)
+    return data 
+
+data = load_data(filepath = "data/train_data/train_table_seq_1.jsonl")
+data += load_data(filepath = "data/train_data/train_table_seq_2.jsonl")
+with open("data/train.jsonl", "w") as f:
+    for item in data:
+        f.write(json.dumps(item) + "\n")
+        

data/format_conversion/iter_gen.py

@@ -0,0 +1,44 @@
+import jsonlines
+
+import os    
+def file_name(file_dir):   
+    L=[]   
+    for root, dirs, files in os.walk(file_dir):  
+        for file in files:  
+           if os.path.splitext(file)[1] == '.jsonl' and "iter" not in os.path.splitext(file)[0] and "tapex" not in os.path.splitext(file)[0]:  
+                L.append(os.path.join(root, file))  
+    return L  
+
+def iter(path):
+    with jsonlines.open(f"{path.replace('.jsonl','_iter.jsonl')}", mode='w') as writer:
+        with jsonlines.open(f"{path}", "r") as f:
+            for item in f:
+                output=item['output'][0].split(' | ')
+                for i in range(len(output)):
+                    if "SCENE DESCRIPTION" not in item['input']:
+                        id=item['id']+f'%{i}'
+                        input=[]
+                        input.append(item['input'])
+                        for _ in range(i):
+                            input.append(output[_])
+                        input=" | ".join(input)
+                        z=[output[i]]
+                        dic={'id':id,"input":input,"output":z}
+                        writer.write(dic)
+                    else :
+                        id=item['id']+f'%{i}'
+                        input_split=item['input'].split("SCENE DESCRIPTION")
+                        scene_info=input_split[1]
+                        input=[]
+                        input.append(input_split[0])
+                        for _ in range(i):
+                            input.append(output[_])
+                        input=" | ".join(input)
+                        z=[output[i]]
+                        dic={'id':id,"input":input+"SCENE DESCRIPTION"+scene_info,"output":z}
+                        writer.write(dic)
+    
+L=file_name('./data')
+for i in L:
+    print(i)
+    iter(i)

data/format_conversion/tapex_gen.py

@@ -0,0 +1,51 @@
+import jsonlines
+import json
+import os    
+def file_name(file_dir):   
+    L=[]   
+    for root, dirs, files in os.walk(file_dir):  
+        for file in files:  
+           if os.path.splitext(file)[1] == ".jsonl" and "iter" not in os.path.splitext(file)[0] and "tapex" not in os.path.splitext(file)[0] and 'table' in os.path.splitext(file)[0]:  
+                L.append(os.path.join(root, file))  
+    return L  
+
+
+attributes=["id","object_type","position_x","position_y","position_z","rotation_x","rotation_y","rotation_z","parent_receptacle"]
+
+def convert(item):
+    input_table={attribute:[] for attribute in attributes}
+    id=item["id"]
+    input_split=item["input"].split("SCENE DESCRIPTION")
+    try:
+        scene_info=input_split[1]
+    except:
+        try:
+            scene_info=item["input"].split(":")[1]
+        except:return False
+    k=scene_info.split("[SEP]")
+    t=True
+    for object in k:
+        if t:
+            t=False
+            continue 
+        atts=object.split(',')
+        for att,attribute in zip(atts,attributes):
+            input_table[attribute].append(att.strip())
+    dic={"id":id,"input":input_split[0],"output":item["output"][0],"input_table":input_table}
+    return dic
+
+
+def convert_file(path):
+    with jsonlines.open(f"{path}", "r") as f:
+        with jsonlines.open(f'{path.replace(".jsonl","_tapex.jsonl")}', mode="w") as writer:
+            for item in f:
+                if convert(item)==False:continue
+                writer.write(convert(item))
+    
+
+
+L=file_name("./data")
+print(L)
+for i in L:
+    print(i)
+    convert_file(i)

data/readme.md

@@ -0,0 +1,9 @@
+This folder contains the G-PlanET dataset and the data scripts used to generate data for iterative and tapex
+
+run
+```
+iter_gen.py
+
+tapex_gen.py
+``` 
+in the folder `format conversion` will generate the data for iterative training and tapex training and testing

data/test_data/tests_seen_table_seq_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df9776e4146bd1e276e300603b42d021d3b3b0bdfa008e0ad15ae02ffa8c4d2a
+size 25883297

data/test_data/tests_unseen_table_seq_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6de7055d49ef98aa7271c91ae6d5b9e075fcf9410657948d881a1c25de499833
+size 24273672

data/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:081b7669bb327ca9ff8e7f5ac72883136dd098c055ec9240cc49f70459807dcc
+size 306598433

data/train_data/train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e765613044dc9b749b24c95bd99722548fe7f5bc03f05277900367faf637f8f5
+size 11746395

data/train_data/train_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0123279a9dfc668c484db9515cdf6876eaaae3691e65ee0eaffc6a65cce0ab2
+size 67672122

data/train_data/train_table_seq_1.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20246be8a050f2f74e18dd4be521399b677099918e45f2448e0d0103370fe4f3
+size 47746372

data/train_data/train_table_seq_1_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6edc50f2949b7ff21e979dcbadc580799f97b9c182815f4acea1b157f734d68a
+size 370402104

data/train_data/train_table_seq_1_tapex.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2216d2734fa9c44d68ebe15768f1840a756cd4d49364d31859a70e70b61b5fb4
+size 58170797

data/train_data/train_table_seq_2.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5f5c24052b8167cae054498d73837c92faa797e1923434ceec2790797025df5
+size 47288736

data/train_data/train_table_seq_2_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:586121c816c020ad2d238e5591b2a9fdc8976627b6bfd401e7385d122f679d35
+size 363377921

data/train_data/train_table_seq_2_tapex.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ecfe56df3310686dbbca101ebb2880094e3c7380c0f88a55e670f7c258a895c
+size 57614866

data/valid_data/valid_seen_table_seq_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cc7ecfd6f62f499d27f1d5d679c54b459342707dc61955014a5dc8b1d8da85c
+size 29341005

data/valid_data/valid_unseen_table_seq_iter.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b61cb2dcadc673c4d627e62fc83db38eef0075f566aea70388c942d65284b81e
+size 28336600