pyg-nightly 2.7.0.dev20241119__py3-none-any.whl → 2.7.0.dev20241120__py3-none-any.whl

{pyg_nightly-2.7.0.dev20241119.dist-info → pyg_nightly-2.7.0.dev20241120.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: pyg-nightly
-Version: 2.7.0.dev20241119
+Version: 2.7.0.dev20241120
 Summary: Graph Neural Network Library for PyTorch
 Keywords: deep-learning,pytorch,geometric-deep-learning,graph-neural-networks,graph-convolutional-networks
 Author-email: Matthias Fey <matthias@pyg.org>

{pyg_nightly-2.7.0.dev20241119.dist-info → pyg_nightly-2.7.0.dev20241120.dist-info}/RECORD

@@ -1,4 +1,4 @@
-torch_geometric/__init__.py,sha256=spxW7Bk1ADYtDbAY5o7hc4aHzY-HMhp_JzJaHacQX30,1904
+torch_geometric/__init__.py,sha256=6TaGIcKhVXKpuS3oTnQXpL4e32Kq0tg7fGKeaWvRoZU,1904
 torch_geometric/_compile.py,sha256=f-WQeH4VLi5Hn9lrgztFUCSrN_FImjhQa6BxFzcYC38,1338
 torch_geometric/_onnx.py,sha256=V9ffrIKSqhDw6xUZ12lkuSfNs48cQp2EeJ6Z19GfnVw,349
 torch_geometric/backend.py,sha256=lVaf7aLoVaB3M-UcByUJ1G4T4FOK6LXAg0CF4W3E8jo,1575
@@ -53,7 +53,7 @@ torch_geometric/data/temporal.py,sha256=WOJ6gFrTLikaLhUvotyUF5ql14FkE5Ox3hNkdSp6
 torch_geometric/data/view.py,sha256=XjkVSc-UWZFCT4DlXLShZtO8duhFQkS9gq88zZXANsk,1089
 torch_geometric/data/lightning/__init__.py,sha256=w3En1tJfy3kSqe1MycpOyZpHFO3fxBCgNCUOznPA3YU,178
 torch_geometric/data/lightning/datamodule.py,sha256=Bn9iaIfE4NWDDWWMqCvBeZ4bIW1Silx_Ol5CPJCliaQ,29242
-torch_geometric/datasets/__init__.py,sha256=fey-955PyCQXGBeUTNPWwU5uK3PJOEvaY1_fDt1SxXc,5880
+torch_geometric/datasets/__init__.py,sha256=f9YqoX9WTSVMzjuLfFD_eCsC4iQ5kbFNQiZru3n6qw0,6013
 torch_geometric/datasets/actor.py,sha256=oUxgJIX8bi5hJr1etWNYIFyVQNDDXi1nyVpHGGMEAGQ,4304
 torch_geometric/datasets/airfrans.py,sha256=212gYsk7PvF-qcmvM2YXaOBhFrS79evAGg_sPHXih4w,5439
 torch_geometric/datasets/airports.py,sha256=b3gkv3gY2JkUpmGiz36Z-g7EcnSfU8lBG1YsCOWdJ6k,3758
@@ -113,6 +113,7 @@ torch_geometric/datasets/md17.py,sha256=BD6LU2xm6_ycXVk6r4O0poNt5Sr_PJ2P1QjNqIOL
 torch_geometric/datasets/mixhop_synthetic_dataset.py,sha256=4NNvTHUvvV6pcqQCyVDS5XhppXUeF2H9GTfFoc49eyU,3951
 torch_geometric/datasets/mnist_superpixels.py,sha256=o2ArbZ0_OE0u8VCaHmWwvngESlOFr9oM9dSEP_tjAS4,3340
 torch_geometric/datasets/modelnet.py,sha256=-qmLjlQiKVWmtHefAIIE97dQxEcaBfetMJnvgYZuwkg,5347
+torch_geometric/datasets/molecule_gpt_dataset.py,sha256=XE14wgPVBm2kVLYL6NgXUDhv4QGHxVISG-VWEwO7hfA,18754
 torch_geometric/datasets/molecule_net.py,sha256=VNWLEDulFID8mLsxgN8q1T-O3M2i0n0Si5ISwEZezMU,7379
 torch_geometric/datasets/movie_lens.py,sha256=M4Bu0Xus8IkW8GYzjxPxSdPXNbcCCx9cu6cncxBvLx8,4033
 torch_geometric/datasets/movie_lens_100k.py,sha256=eTpBAteM3jqTEtiwLxmhVj4r8JvftvPx8Hvs-3ZIHlU,6057
@@ -144,6 +145,7 @@ torch_geometric/datasets/shapenet.py,sha256=tn3HiQQAr6lxHrqxfOVaAtl40guwFYTXWCbS
 torch_geometric/datasets/shrec2016.py,sha256=cTLhctbqE0EUEvKddJFhPzDb1oLKXOth4O_WzsWtyMk,6323
 torch_geometric/datasets/snap_dataset.py,sha256=r3sC-dHDouyaYoHGdoBY0uO0qOOvD6_Hb96d2ceGMZk,9433
 torch_geometric/datasets/suite_sparse.py,sha256=eqjH4vAUq872qdk3YdLkZSwlu6r7HHpTgK0vEVGmY1s,3278
+torch_geometric/datasets/tag_dataset.py,sha256=0fzOsakR9L9CK6ppGN-USD4-Vq-ssbQ2Xovw2nqqtWo,14759
 torch_geometric/datasets/taobao.py,sha256=CUcZpbWsNTasevflO8zqP0YvENy89P7wpKS4MHaDJ6Q,4170
 torch_geometric/datasets/tosca.py,sha256=nUSF8NQT1GlkwWQLshjWmr8xORsvRHzzIqhUyDCvABc,4632
 torch_geometric/datasets/tu_dataset.py,sha256=14OSaXBgVwT1dX2h1wZ3xVIwoo0GQBEfR3yWh6Q0VF0,7847
@@ -324,8 +326,9 @@ torch_geometric/nn/aggr/set_transformer.py,sha256=FG7_JizpFX14M6VSCwLSjYXYdJ1ZiQ
 torch_geometric/nn/aggr/sort.py,sha256=bvOOWnFkNOBOZih4rqVZQsjfeDX3vmXo1bpPSFD846w,2507
 torch_geometric/nn/aggr/utils.py,sha256=CLJ-ZrVWYIOBpdhQBLAz94dj3cMKKKc3qwGr4DFbiCU,8338
 torch_geometric/nn/aggr/variance_preserving.py,sha256=fu-U_aGYpVLpgSFvVg0ONMe6nqoyv8tZ6Y35qMYTf9w,1126
-torch_geometric/nn/attention/__init__.py,sha256=Ip6n4xbUbhJhrmPO9LjvHq0nNQe-yxiC4WHyOYOrHJc,76
+torch_geometric/nn/attention/__init__.py,sha256=1lCB7zh7uM6FkpW81S9U4CvxTwpCkz59KatPTIE9UmA,127
 torch_geometric/nn/attention/performer.py,sha256=2PCDn4_-oNTao2-DkXIaoi18anP01OxRELF2pvp-jk8,7357
+torch_geometric/nn/attention/qformer.py,sha256=7J-pWm_vpumK38IC-iCBz4oqL-BEIofEIxJ0wfjWq9A,2338
 torch_geometric/nn/conv/__init__.py,sha256=37zTdt0gfSAUPMtwXjZg5mWx_itojJVFNODYR1h1ch0,3515
 torch_geometric/nn/conv/agnn_conv.py,sha256=5nEPLx_BBHcDaO6HWzLuHfXc0Yd_reKynAOH0Iq09lU,3077
 torch_geometric/nn/conv/antisymmetric_conv.py,sha256=dhA6sCETy1jlXReYJZBSyToOcL_mZ1wL10fMIb8Ppuw,4387
@@ -417,7 +420,7 @@ torch_geometric/nn/kge/distmult.py,sha256=dGQ0bVzjreZgFN1lXE23_IIidsiOq7ehPrMb-N
 torch_geometric/nn/kge/loader.py,sha256=5Uc1j3OUMQnBYSHDqL7pLCty1siFLzoPkztigYO2zP8,771
 torch_geometric/nn/kge/rotate.py,sha256=XLuO1AbyTt5cJxr97ZzoyAyIEsHKesgW5TvDmnGJAao,3208
 torch_geometric/nn/kge/transe.py,sha256=jlejq5BLMm-sb1wWcLDp7pZqCdelWBgjDIC8ctbjSdU,3088
-torch_geometric/nn/models/__init__.py,sha256=RpYFFqaYWq1BVMF3Fs-EQo-QZDdLQjIHPdkl3d2MOW4,2017
+torch_geometric/nn/models/__init__.py,sha256=dr2-YsRzUdVBM6Ut78FB9Wbjn-kzV0gPwOlWGPdQLY4,2108
 torch_geometric/nn/models/attentive_fp.py,sha256=tkgvw28wg9-JqHIfBllfCwTHrZIUiv85yZJcDqjz3z0,6634
 torch_geometric/nn/models/autoencoder.py,sha256=nGje-zty78Y3hxOJ9o0_6QziJjOvBlknk6z0_fDQwQU,10770
 torch_geometric/nn/models/basic_gnn.py,sha256=PGa0RUMyvrNy_5yRI2jX_zwPsmZXwOQWfsWvxOiHsSk,31225
@@ -428,6 +431,7 @@ torch_geometric/nn/models/deepgcn.py,sha256=tIgT03cj8MghYlxEozpoGvGG_CwpJrGDxv1Z
 torch_geometric/nn/models/dimenet.py,sha256=Kc5p-rB5q-0e8lY22l-OdQTscTxJh2lTEpeRFMdL4RY,36186
 torch_geometric/nn/models/dimenet_utils.py,sha256=Eyn_EiJqwKvuYj6BtRpSxrzMG3v4Gk98X9MxZ7uvwm4,5069
 torch_geometric/nn/models/g_retriever.py,sha256=VueRImNJlh1WvRWcsSXliSw8RlxlzWlu2WSFs_VQaJc,7749
+torch_geometric/nn/models/glem.py,sha256=gqQF4jlU7U_u5-zGeJZuHiEqhSXa-wLU5TghN4u5fYY,16389
 torch_geometric/nn/models/gnnff.py,sha256=15dkiLgy0LmH1hnUrpeoHioIp4BPTfjpVATpnGRt9E0,7860
 torch_geometric/nn/models/graph_mixer.py,sha256=mthMeCOikR8gseEsu4oJ3Cd9C35zHSv1p32ROwnG-6s,9246
 torch_geometric/nn/models/graph_unet.py,sha256=N8TSmJo8AlbZjjcame0xW_jZvMOirL5ahw6qv5Yjpbs,5586
@@ -439,6 +443,7 @@ torch_geometric/nn/models/mask_label.py,sha256=B2HcL6ZkaUEo3a8nebZoUqEIfDEfcIGOV
 torch_geometric/nn/models/meta.py,sha256=lQWovjdQgTGT_rDAm6L186ObINeQCD9tLBz8xenmrF0,6540
 torch_geometric/nn/models/metapath2vec.py,sha256=nxttGe4QVWr4teYEoNz8uHRu-yVsLSZPOeF_tz0bj2o,10788
 torch_geometric/nn/models/mlp.py,sha256=rdwUFxxxqLjXK-iy1L1sXiwSNwAfqTlvHLaqVZ-jwCs,10315
+torch_geometric/nn/models/molecule_gpt.py,sha256=k-XULH6jaurj-R2EE4sIWTkqlNqa3CzWxfQgfFa-G8s,7637
 torch_geometric/nn/models/neural_fingerprint.py,sha256=pTLJgU9Uh2Lnf9bggLj4cKI8YdEFcMF-9MALuubqbuQ,2378
 torch_geometric/nn/models/node2vec.py,sha256=U-VhJlvt5lT-JShFrF5tN84wCPqoVuftLVNyOVXs0OU,7664
 torch_geometric/nn/models/pmlp.py,sha256=dcAASVSyQMMhItSfEJWPeAFh0R3tNCwAHwdrShwQ8o4,3538
@@ -450,8 +455,8 @@ torch_geometric/nn/models/signed_gcn.py,sha256=J40CnedFIqtKI1LhW1ITSEFRbA_XiJZL6
 torch_geometric/nn/models/tgn.py,sha256=kEGdfLJybkbMT4UMoAh2nCzfX3_nDjfm1cicuPHEwAM,11878
 torch_geometric/nn/models/visnet.py,sha256=97OFMCsPDEI5BCSi7RhoRcU2CNRp7zck2tEzrltFZj4,43192
 torch_geometric/nn/nlp/__init__.py,sha256=JJESTA7w_K8v60XbCd25IqmrKKHLz5OiNexMHYGV2mE,138
-torch_geometric/nn/nlp/llm.py,sha256=_penl2qkDMeVtlwGPrl7UuyxBh6ILtdiLHmrUNQHkYc,11731
-torch_geometric/nn/nlp/sentence_transformer.py,sha256=JrTN3W1srdkNX7qYDGB08mY5615i5nfEJSTHAdd5EuA,3260
+torch_geometric/nn/nlp/llm.py,sha256=M15Qn0yHyA6HL2rHCH2p4H6hKjUvLfnzlxdfEFvRxSA,11732
+torch_geometric/nn/nlp/sentence_transformer.py,sha256=VzMtNUYk6FvOVc3PdVets9_2Sb2FdQbzu9H3m6teRlI,3417
 torch_geometric/nn/norm/__init__.py,sha256=u2qIDrkbeuObGVXSAIftAlvSd6ouGTtxznCfD-59UiA,669
 torch_geometric/nn/norm/batch_norm.py,sha256=sJKrinHGwA-noIgteg1RD2W06rd0zskD-rXuY-36glY,8283
 torch_geometric/nn/norm/diff_group_norm.py,sha256=b57XvNekrUYGDjNJlGeqvaMGNJmHwopSF0_yyBWlLuA,4722
@@ -618,6 +623,6 @@ torch_geometric/utils/undirected.py,sha256=H_nfpI0_WluOG6VfjPyldvcjL4w5USAKWu2x5
 torch_geometric/visualization/__init__.py,sha256=PyR_4K5SafsJrBr6qWrkjKr6GBL1b7FtZybyXCDEVwY,154
 torch_geometric/visualization/graph.py,sha256=ZuLPL92yGRi7lxlqsUPwL_EVVXF7P2kMcveTtW79vpA,4784
 torch_geometric/visualization/influence.py,sha256=CWMvuNA_Nf1sfbJmQgn58yS4OFpeKXeZPe7kEuvkUBw,477
-pyg_nightly-2.7.0.dev20241119.dist-info/WHEEL,sha256=CpUCUxeHQbRN5UGRQHYRJorO5Af-Qy_fHMctcQ8DSGI,82
-pyg_nightly-2.7.0.dev20241119.dist-info/METADATA,sha256=3Y-GTdZXsDzzOxIrxa35EwttZ_dPAwF2jkLotBJ9ubg,62979
-pyg_nightly-2.7.0.dev20241119.dist-info/RECORD,,
+pyg_nightly-2.7.0.dev20241120.dist-info/WHEEL,sha256=CpUCUxeHQbRN5UGRQHYRJorO5Af-Qy_fHMctcQ8DSGI,82
+pyg_nightly-2.7.0.dev20241120.dist-info/METADATA,sha256=vWh7yN2ZZvIhSw04tE8VdrXEQZ5TFeItmVYp8xPL7kQ,62979
+pyg_nightly-2.7.0.dev20241120.dist-info/RECORD,,

torch_geometric/__init__.py

@@ -30,7 +30,7 @@ from .lazy_loader import LazyLoader
 contrib = LazyLoader('contrib', globals(), 'torch_geometric.contrib')
 graphgym = LazyLoader('graphgym', globals(), 'torch_geometric.graphgym')
 
-__version__ = '2.7.0.dev20241119'
+__version__ = '2.7.0.dev20241120'
 
 __all__ = [
     'Index',

torch_geometric/datasets/__init__.py

@@ -77,6 +77,8 @@ from .myket import MyketDataset
 from .brca_tgca import BrcaTcga
 from .neurograph import NeuroGraphDataset
 from .web_qsp_dataset import WebQSPDataset
+from .molecule_gpt_dataset import MoleculeGPTDataset
+from .tag_dataset import TAGDataset
 
 from .dbp15k import DBP15K
 from .aminer import AMiner
@@ -190,6 +192,8 @@ homo_datasets = [
     'BrcaTcga',
     'NeuroGraphDataset',
     'WebQSPDataset',
+    'MoleculeGPTDataset',
+    'TAGDataset',
 ]
 
 hetero_datasets = [

torch_geometric/datasets/molecule_gpt_dataset.py

@@ -0,0 +1,480 @@
+import gzip
+import json
+import multiprocessing
+import os
+import sys
+from collections import defaultdict
+from multiprocessing import Pool
+from typing import Callable, List, Optional, Tuple
+
+import numpy as np
+import requests
+import torch
+from tqdm import tqdm
+
+from torch_geometric.data import Data, InMemoryDataset, download_url
+from torch_geometric.io import fs
+from torch_geometric.nn.nlp import LLM
+from torch_geometric.utils import one_hot
+
+
+def clean_up_description(description: str) -> str:
+    description = description + " "
+
+    # extra adj Pure
+    if description.startswith("Pure "):
+        description = description.replace("Pure ", "")
+    # fix typo
+    if description.startswith("Mercurycombines"):
+        description = description.replace("Mercurycombines",
+                                          "Mercury combines")
+
+    # a special case
+    description = description.replace(
+        "17-Hydroxy-6-methylpregna-3,6-diene-3,20-dione. ",
+        "17-Hydroxy-6-methylpregna-3,6-diene-3,20-dione is ")
+
+    # a special case
+    description = description.replace("5-Thymidylic acid. ",
+                                      "5-Thymidylic acid. is ")
+
+    # a special case
+    description = description.replace(
+        "5'-S-(3-Amino-3-carboxypropyl)-5'-thioadenosine. ",
+        "5'-S-(3-Amino-3-carboxypropyl)-5'-thioadenosine. is ")
+
+    # a special case
+    description = description.replace(
+        ("Guanosine 5'-(trihydrogen diphosphate), monoanhydride"
+         " with phosphorothioic acid. "),
+        ("Guanosine 5'-(trihydrogen diphosphate), monoanhydride"
+         " with phosphorothioic acid is "))
+
+    # a special case
+    description = description.replace("5'-Uridylic acid. ",
+                                      "5'-Uridylic acid is ")
+
+    # a special case
+    description = description.replace("5'-Adenylic acid, ",
+                                      "5'-Adenylic acid is ")
+
+    # a special case
+    description = description.replace(
+        "Uridine 5'-(tetrahydrogen triphosphate). ",
+        "Uridine 5'-(tetrahydrogen triphosphate). is ")
+
+    # a special case
+    description = description.replace("Inosine 5'-Monophosphate. ",
+                                      "Inosine 5'-Monophosphate. is ")
+
+    # a special case
+    description = description.replace("Pivaloyloxymethyl butyrate (AN-9), ",
+                                      "Pivaloyloxymethyl butyrate (AN-9) is ")
+
+    # a special case
+    description = description.replace(
+        "4-Amino-5-cyano-7-(D-ribofuranosyl)-7H- pyrrolo(2,3-d)pyrimidine. ",
+        "4-Amino-5-cyano-7-(D-ribofuranosyl)-7H- pyrrolo(2,3-d)pyrimidine is ")
+
+    # a special case
+    description = description.replace(
+        "Cardamonin (also known as Dihydroxymethoxychalcone), ",
+        "Cardamonin (also known as Dihydroxymethoxychalcone) is ")
+
+    # a special case
+    description = description.replace("Lithium has been used to treat ",
+                                      "Lithium is ")
+
+    # a special case
+    description = description.replace("4,4'-Methylenebis ",
+                                      "4,4'-Methylenebis is ")
+
+    # a special case
+    description = description.replace(
+        "2,3,7,8-Tetrachlorodibenzo-p-dioxin",
+        "2,3,7,8-Tetrachlorodibenzo-p-dioxin is ")
+
+    # a special case
+    description = description.replace("Exposure to 2,4,5-trichlorophenol ",
+                                      "2,4,5-Trichlorophenol exposure ")
+
+    index = 0
+    L = len(description)
+    if description.startswith('C.I. '):
+        start_index = len('C.I. ')
+    elif description.startswith('Nectriapyrone. D '):
+        start_index = len('Nectriapyrone. D ')
+    elif description.startswith(
+            'Salmonella enterica sv. Minnesota LPS core oligosaccharide'):
+        start_index = len(
+            'Salmonella enterica sv. Minnesota LPS core oligosaccharide')
+    else:
+        start_index = 0
+    for index in range(start_index, L - 1):
+        if index < L - 2:
+            if description[index] == '.' and description[
+                    index + 1] == ' ' and 'A' <= description[index + 2] <= 'Z':
+                break
+        elif index == L - 2:
+            break
+
+    first_sentence = description[:index + 1]
+    return first_sentence
+
+
+def extract_name(name_raw: str, description: str) -> Tuple[str, str, str]:
+    first_sentence = clean_up_description(description)
+
+    splitter = '  --  --  '
+    if ' are ' in first_sentence or ' were ' in first_sentence:
+        replaced_words = 'These molecules'
+    else:
+        replaced_words = 'This molecule'
+
+    first_sentence = first_sentence.replace(' is ', splitter)
+    first_sentence = first_sentence.replace(' are ', splitter)
+    first_sentence = first_sentence.replace(' was ', splitter)
+    first_sentence = first_sentence.replace(' were ', splitter)
+    first_sentence = first_sentence.replace(' appears ', splitter)
+    first_sentence = first_sentence.replace(' occurs ', splitter)
+    first_sentence = first_sentence.replace(' stands for ', splitter)
+    first_sentence = first_sentence.replace(' belongs to ', splitter)
+    first_sentence = first_sentence.replace(' exists ',
+                                            splitter)  # only for CID=11443
+    first_sentence = first_sentence.replace(' has been used in trials ',
+                                            splitter)
+    first_sentence = first_sentence.replace(' has been investigated ',
+                                            splitter)
+    first_sentence = first_sentence.replace(' has many uses ', splitter)
+
+    if splitter in first_sentence:
+        extracted_name = first_sentence.split(splitter, 1)[0]
+    elif first_sentence.startswith(name_raw):
+        extracted_name = name_raw
+    elif name_raw in first_sentence:
+        extracted_name = name_raw
+        extracted_name = None
+        print("=====", name_raw)
+        print("first sentence: ", first_sentence)
+    else:
+        extracted_name = None
+
+    if extracted_name is not None:
+        extracted_description = description.replace(extracted_name,
+                                                    replaced_words)
+    else:
+        extracted_description = description
+
+    return extracted_name, extracted_description, first_sentence
+
+
+class MoleculeGPTDataset(InMemoryDataset):
+    r"""The dataset from the `"MoleculeGPT: Instruction Following Large
+    Language Models for Molecular Property Prediction"
+    <https://ai4d3.github.io/papers/34.pdf>`_ paper.
+
+    Args:
+        root (str): Root directory where the dataset should be saved.
+        transform (callable, optional): A function/transform that takes in an
+            :obj:`torch_geometric.data.Data` object and returns a transformed
+            version. The data object will be transformed before every access.
+            (default: :obj:`None`)
+        pre_transform (callable, optional): A function/transform that takes in
+            an :obj:`torch_geometric.data.Data` object and returns a
+            transformed version. The data object will be transformed before
+            being saved to disk. (default: :obj:`None`)
+        pre_filter (callable, optional): A function that takes in an
+            :obj:`torch_geometric.data.Data` object and returns a boolean
+            value, indicating whether the data object should be included in the
+            final dataset. (default: :obj:`None`)
+        force_reload (bool, optional): Whether to re-process the dataset.
+            (default: :obj:`False`)
+        total_page_num (int, optional): The number of pages from PubChem.
+            (default: :obj:`10`)
+        total_block_num (int, optional): The blocks of SDF files from PubChem.
+            (default: :obj:`1`)
+    """
+    description_url = (
+        'https://pubchem.ncbi.nlm.nih.gov/rest/pug_view/annotations/'
+        'heading/json?heading_type=Compound&heading=Record+Description&page={}'
+    )
+    compound_url = ('https://ftp.ncbi.nlm.nih.gov/pubchem/Compound/'
+                    'CURRENT-Full/SDF')
+
+    def __init__(
+        self,
+        root: str,
+        transform: Optional[Callable] = None,
+        pre_transform: Optional[Callable] = None,
+        pre_filter: Optional[Callable] = None,
+        force_reload: bool = False,
+        total_page_num: int = 10,
+        total_block_num: int = 1,
+    ):
+        self.total_page_num = total_page_num
+        self.total_block_num = total_block_num
+
+        super().__init__(root, transform, pre_transform, pre_filter,
+                         force_reload=force_reload)
+        self.load(self.processed_paths[0])
+
+    @property
+    def raw_file_names(self) -> List[str]:
+        return ['pubchem.csv']
+
+    @property
+    def processed_file_names(self) -> List[str]:
+        return ['data.pt']
+
+    def download(self) -> None:
+        # Step 01. Extract description
+        step1_folder = f"{self.raw_dir}/step_01_PubChemSTM_description"
+        if not os.path.exists(step1_folder):
+            os.makedirs(step1_folder)
+            valid_CID_set = set()
+            CID2name_raw, CID2name_extracted = defaultdict(list), defaultdict(
+                list)
+            CID2text_raw, CID2text_extracted = defaultdict(list), defaultdict(
+                list)
+
+            for page_index in tqdm(range(self.total_page_num)):
+                page_num = page_index + 1
+                f_out = open(
+                    f"{step1_folder}/Compound_description_{page_num}.txt", "w")
+
+                description_data = requests.get(
+                    self.description_url.format(page_num)).json()
+
+                description_data = description_data["Annotations"]
+                assert description_data["Page"] == page_num
+
+                record_list = description_data["Annotation"]
+
+                for record in record_list:
+                    try:
+                        CID = record["LinkedRecords"]["CID"][0]
+                        if "Name" in record:
+                            name_raw = record["Name"]
+                            CID2name_raw[CID].append(name_raw)
+                        else:
+                            name_raw = None
+
+                        data_list = record["Data"]
+                        for data in data_list:
+                            description = data["Value"]["StringWithMarkup"][0][
+                                "String"].strip()
+
+                            extracted_name, extracted_description, _ = extract_name(  # noqa: E501
+                                name_raw, description)
+                            if extracted_name is not None:
+                                CID2name_extracted[CID].append(extracted_name)
+
+                            CID2text_raw[CID].append(description)
+                            CID2text_extracted[CID].append(
+                                extracted_description)
+
+                            valid_CID_set.add(CID)
+                            f_out.write(f"{CID}\n")
+                            f_out.write(f"{extracted_description}\n\n")
+                    except Exception:
+                        continue
+
+            valid_CID_list = sorted(list(valid_CID_set))
+            print(f"Total CID (with raw name) {len(CID2name_raw)}")
+            print(f"Total CID (with extracted name) {len(CID2name_extracted)}")
+            print(f"Total CID {len(valid_CID_list)}")
+
+            with open(f"{self.raw_dir}/CID2name_raw.json", "w") as f:
+                json.dump(CID2name_raw, f)
+
+            with open(f"{self.raw_dir}/CID2name.json", "w") as f:
+                json.dump(CID2name_extracted, f)
+
+            with open(f"{self.raw_dir}/CID2text_raw.json", "w") as f:
+                json.dump(CID2text_raw, f)
+
+            with open(f"{self.raw_dir}/CID2text.json", "w") as f:
+                json.dump(CID2text_extracted, f)
+
+        # Step 02. Download SDF Files
+        step2_folder = f"{self.raw_dir}/step_02_PubChemSTM_SDF"
+        if not os.path.exists(step2_folder):
+            for block_id in tqdm(range(self.total_block_num)):
+                block_size = 500000
+                l_id = block_id * block_size + 1
+                r_id = (block_id + 1) * block_size
+
+                compound_file_name = f"Compound_{l_id:09d}_{r_id:09d}.sdf.gz"
+                download_url(f"{self.compound_url}/{compound_file_name}",
+                             step2_folder)
+
+    def process(self, use_mp: bool = False) -> None:
+        try:
+            from rdkit import Chem
+            from rdkit.Chem.rdchem import BondType as BT
+            WITH_RDKIT = True
+
+        except ImportError:
+            WITH_RDKIT = False
+
+        if not WITH_RDKIT:
+            print(("Using a pre-processed version of the dataset. Please "
+                   "install 'rdkit' to alternatively process the raw data."),
+                  file=sys.stderr)
+
+            data_list = fs.torch_load(self.raw_paths[0])
+            data_list = [Data(**data_dict) for data_dict in data_list]
+
+            if self.pre_filter is not None:
+                data_list = [d for d in data_list if self.pre_filter(d)]
+
+            if self.pre_transform is not None:
+                data_list = [self.pre_transform(d) for d in data_list]
+
+            self.save(data_list, self.processed_paths[0])
+            return
+
+        # Step 03. Filter out SDF
+        step2_folder = f"{self.raw_dir}/step_02_PubChemSTM_SDF"
+        step3_folder = f"{self.raw_dir}/step_03_PubChemSTM_filtered"
+        if not os.path.exists(step3_folder):
+            os.makedirs(step3_folder)
+            with open(f"{self.raw_dir}/CID2text.json") as f:
+                CID2text = json.load(f)
+            target_CID_list = set(CID2text.keys())
+
+            block_size = 500000
+
+            def extract_one_SDF_file(block_id: int) -> None:
+                valid_mol_count = 0
+
+                writer = Chem.SDWriter(
+                    f'{step3_folder}/filtered_{block_id}.sdf')
+                l_id = block_id * block_size + 1
+                r_id = (block_id + 1) * block_size
+
+                compound_file_name = f"Compound_{l_id:09d}_{r_id:09d}.sdf.gz"
+                gzip_loader = gzip.open(f"{step2_folder}/{compound_file_name}")
+                suppl = Chem.ForwardSDMolSupplier(gzip_loader)
+
+                for mol in tqdm(suppl):
+                    if mol is None:
+                        continue
+                    cid = mol.GetProp("PUBCHEM_COMPOUND_CID")
+
+                    if cid not in target_CID_list:
+                        continue
+
+                    writer.write(mol)
+                    valid_mol_count += 1
+
+                print(f"block id: {block_id}\nfound {valid_mol_count}\n\n")
+                sys.stdout.flush()
+                return
+
+            if use_mp:
+                num_process = multiprocessing.cpu_count()
+                print(f"{num_process} CPUs")
+                num_process = 8
+                p = Pool(num_process)
+
+                block_id_list = np.arange(self.total_block_num)
+                with p:
+                    p.map(extract_one_SDF_file, block_id_list)
+            else:
+                for block_id in range(self.total_block_num):
+                    extract_one_SDF_file(block_id)
+
+        # Step 04. Merge SDF
+        with open(f"{self.raw_dir}/CID2text.json") as f:
+            CID2text = json.load(f)
+        target_CID_list = set(CID2text.keys())
+        print(f'The length of target_CID_list: {len(target_CID_list)}')
+
+        writer = Chem.SDWriter(f'{self.raw_dir}/molecules.sdf')
+
+        found_CID_set = set()
+        for block_id in range(self.total_block_num + 1):
+            compound_file_path = f"{step3_folder}/filtered_{block_id}.sdf"
+            try:
+                suppl = Chem.SDMolSupplier(compound_file_path)
+
+                for mol in tqdm(suppl):
+                    writer.write(mol)
+                    cid = mol.GetProp("PUBCHEM_COMPOUND_CID")
+                    found_CID_set.add(cid)
+            except Exception:
+                print(f"block id: {block_id} with 0 valid SDF file")
+                continue
+
+        print(f"In total: {len(found_CID_set)} molecules")
+
+        # Step 05. Convert to PyG data format
+        types = {'H': 0, 'C': 1, 'N': 2, 'O': 3, 'F': 4, 'Unknow': 5}
+        bonds = {BT.SINGLE: 0, BT.DOUBLE: 1, BT.TRIPLE: 2, BT.AROMATIC: 3}
+
+        data_list = []
+        # Real data
+        CID2text_file = f'{self.raw_dir}/CID2text.json'
+
+        with open(CID2text_file) as f:
+            CID2text_data = json.load(f)
+
+        suppl = Chem.SDMolSupplier(f'{self.raw_dir}/molecules.sdf')
+
+        llm = LLM(
+            # model_name='lmsys/vicuna-7b-v1.5',
+            model_name='TinyLlama/TinyLlama-1.1B-Chat-v0.1',
+            num_params=1,
+            dtype=torch.bfloat16,
+        )
+        prompt = ("Propose a question regarding the molecule '∼' "
+                  "whose answer is: {}:")
+        for mol in tqdm(suppl):
+            if mol.HasProp('PUBCHEM_COMPOUND_CID'):
+                CID = mol.GetProp("PUBCHEM_COMPOUND_CID")
+                CAN_SMILES = mol.GetProp("PUBCHEM_OPENEYE_CAN_SMILES")
+
+                m: Chem.Mol = Chem.MolFromSmiles(CAN_SMILES)
+                if m is None:
+                    continue
+                RDKit_CAN_SMILES = Chem.MolToSmiles(m)
+
+                ground_truth = CID2text_data[CID][0]
+
+                instruction = llm.inference([prompt.format(ground_truth)])[0]
+
+                x: torch.Tensor = torch.tensor([
+                    types[atom.GetSymbol()] if atom.GetSymbol() in types else 5
+                    for atom in m.GetAtoms()  # type: ignore
+                ])
+                x = one_hot(x, num_classes=len(types), dtype=torch.float)
+
+                rows, cols, edge_types = [], [], []
+                for bond in m.GetBonds():  # type: ignore
+                    i, j = bond.GetBeginAtomIdx(), bond.GetEndAtomIdx()
+                    edge_types += [bonds[bond.GetBondType()]] * 2
+                    rows += [i, j]
+                    cols += [j, i]
+
+                edge_index = torch.tensor([rows, cols], dtype=torch.long)
+                edge_type = torch.tensor(edge_types, dtype=torch.long)
+                edge_attr = one_hot(edge_type, num_classes=len(bonds))
+
+                data = Data(
+                    x=x,
+                    edge_index=edge_index,
+                    edge_attr=edge_attr,
+                    smiles=RDKit_CAN_SMILES,
+                    instruction=instruction,
+                    y=ground_truth,
+                )
+
+                if self.pre_filter is not None and not self.pre_filter(data):
+                    continue
+                if self.pre_transform is not None:
+                    data = self.pre_transform(data)
+
+                data_list.append(data)
+
+        self.save(data_list, self.processed_paths[0])