Upload 12 files

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+dataset/fb15k237/entity2textlong.txt filter=lfs diff=lfs merge=lfs -text

app.py

@@ -1,17 +1,261 @@
 import gradio as gr
+from collections import defaultdict
+from transformers import BertTokenizer, BertForMaskedLM
+import jsonlines
+import torch
+from src.modeling_bert import EXBertForMaskedLM
+from higher.patch import monkeypatch as make_functional
+# from src.models.one_shot_learner import OneShotLearner
 
-def edit_process(title, context):
-    return f"Title:{title}\nContext:{context}\n...", f"Title:{title}\nContext:{context}\n..."
+### load KGE model
+edit_origin_model = BertForMaskedLM.from_pretrained(pretrained_model_name_or_path="zjunlp/KGEditor", subfolder="E-FB15k237")
+edit_ex_model = EXBertForMaskedLM.from_pretrained(pretrained_model_name_or_path="zjunlp/KGEditor", subfolder="E-FB15k237")
+
+edit_learner = torch.load("./learner_checkpoint/edit/learner_params.pt", map_location=torch.device('cpu'))
+add_learner = torch.load("./learner_checkpoint/add/learner_params.pt", map_location=torch.device('cpu'))
+
+add_origin_model = BertForMaskedLM.from_pretrained(pretrained_model_name_or_path="zjunlp/KGEditor", subfolder="A-FB15k237")
+add_ex_model = EXBertForMaskedLM.from_pretrained(pretrained_model_name_or_path="zjunlp/KGEditor", subfolder="A-FB15k237")
+
+### init inputs
+ent_name2id = defaultdict(str)
+id2ent_name = defaultdict(str)
+rel_name2id = defaultdict(str)
+id2ent_text = defaultdict(str)
+id2rel_text = defaultdict(str)
+corrupt_triple = defaultdict(list)
+
+### init tokenizer
+tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
+add_tokenizer = BertTokenizer.from_pretrained(pretrained_model_name_or_path='zjunlp/KGEditor', subfolder="E-FB15k237")
+
+def init_triple_input():
+    global ent2token
+    global ent2id
+    global id2ent
+    global rel2token
+
+    with open("./dataset/fb15k237/relations.txt", "r") as f:
+        lines = f.readlines()
+        relations = []
+        for line in lines:
+            relations.append(line.strip().split('\t')[0])
+
+        rel2token = {ent: f"[RELATION_{i}]" for i, ent in enumerate(relations)}
+        
+    with open("./dataset/fb15k237/entity2text.txt", "r") as f:
+        for line in f.readlines():
+            id, name = line.rstrip('\n').split('\t')
+            ent_name2id[name] = id
+            id2ent_name[id] = name
+
+    with open("./dataset/fb15k237/relation2text.txt", "r") as f:
+        for line in f.readlines():
+            id, name = line.rstrip('\n').split('\t')
+            rel_name2id[name] = id
+            id2rel_text[id] = name
+
+    with open("./dataset/fb15k237/entity2textlong.txt", "r") as f:
+        for line in f.readlines():
+            id, text = line.rstrip('\n').split('\t')
+            id2ent_text[id] = text.replace("\\n", " ").replace("\\", "")
+
+        entities = list(id2ent_text.keys())
+        ent2token = {ent: f"[ENTITY_{i}]" for i, ent in enumerate(entities)}
+        ent2id = {ent: i for i, ent in enumerate(entities)}
+        id2ent = {i: ent for i, ent in enumerate(entities)}
+
+    with jsonlines.open("./dataset/fb15k237/edit_test.jsonl") as f:
+        lines = []
+        for d in f:
+            corrupt_triple[" ".join(d["ori"])] = d["cor"]
+
+def solve(triple, alter_label, edit_task):
+    h, r, t = triple.split("|")
+    if h == "[MASK]":
+        text_a = "[MASK]"
+        text_b = id2rel_text[r] + " " + rel2token[r]
+        text_c = ent2token[ent_name2id[t]] + " " + id2ent_text[ent_name2id[t]]
+        origin_label = corrupt_triple[" ".join([ent_name2id[alter_label], r, ent_name2id[t]])][0] if edit_task else ent_name2id[alter_label]
+    else:
+        text_a = ent2token[ent_name2id[h]]
+        # text_b = id2rel_text[r] + "[PAD]"
+        text_b = id2rel_text[r] + " " + rel2token[r]
+        text_c = "[MASK]" + " " + id2ent_text[ent_name2id[h]]
+        origin_label = corrupt_triple[" ".join([ent_name2id[h], r, ent_name2id[alter_label]])][2] if edit_task else ent_name2id[alter_label]
+
+    if text_a == "[MASK]":
+        input_text_a = tokenizer.sep_token.join(["[MASK]", id2rel_text[r] + "[PAD]"])
+        input_text_b = "[PAD]" + " " + id2ent_text[ent_name2id[t]]
+    else:
+        input_text_a = "[PAD] "
+        input_text_b = tokenizer.sep_token.join([id2rel_text[r] + "[PAD]", "[MASK]" + " " + id2ent_text[ent_name2id[h]]])
+    
+    cond_inputs_text = "{} >> {} || {}".format(
+        add_tokenizer.added_tokens_decoder[ent2id[origin_label] + len(tokenizer)],
+        add_tokenizer.added_tokens_decoder[ent2id[ent_name2id[alter_label]] + len(tokenizer)],
+        input_text_a + input_text_b
+    )
+
+    inputs = tokenizer(
+        f"{text_a} [SEP] {text_b} [SEP] {text_c}",
+        truncation="longest_first",
+        max_length=64,
+        padding="longest",
+        add_special_tokens=True,
+    )
+
+    edit_inputs = tokenizer(
+        input_text_a,
+        input_text_b,
+        truncation="longest_first",
+        max_length=64,
+        padding="longest",
+        add_special_tokens=True,
+    )
+
+    cond_inputs = tokenizer(
+        cond_inputs_text,
+        truncation=True,
+        max_length=64,
+        padding="max_length",
+        add_special_tokens=True,
+    )
+
+    inputs = {
+        "input_ids": torch.tensor(inputs["input_ids"]).unsqueeze(dim=0),
+        "attention_mask": torch.tensor(inputs["attention_mask"]).unsqueeze(dim=0),
+        "token_type_ids": torch.tensor(inputs["token_type_ids"]).unsqueeze(dim=0)
+    }
+
+    edit_inputs = {
+        "input_ids": torch.tensor(edit_inputs["input_ids"]).unsqueeze(dim=0),
+        "attention_mask": torch.tensor(edit_inputs["attention_mask"]).unsqueeze(dim=0),
+        "token_type_ids": torch.tensor(edit_inputs["token_type_ids"]).unsqueeze(dim=0)
+    }
+
+    cond_inputs = {
+        "input_ids": torch.tensor(cond_inputs["input_ids"]).unsqueeze(dim=0),
+        "attention_mask": torch.tensor(cond_inputs["attention_mask"]).unsqueeze(dim=0),
+        "token_type_ids": torch.tensor(cond_inputs["token_type_ids"]).unsqueeze(dim=0)
+    }
+
+    return inputs, cond_inputs, edit_inputs
+
+def get_logits_orig_params_dict(inputs, cond_inputs, alter_label, ex_model, learner):
+    with torch.enable_grad():
+        logits = ex_model.eval()(
+            input_ids=inputs["input_ids"],
+            attention_mask=inputs["attention_mask"],
+        ).logits
+        # print(logits.shape)
+        # logits_orig, logit_for_grad, _ = logits.split([
+        #     len(inputs["input_ids"]) - 1,
+        #     1,
+        #     0,
+        # ])
+        input_ids = inputs['input_ids']
+        _, mask_idx = (input_ids == tokenizer.mask_token_id).nonzero(as_tuple=True)
+        mask_logits = logits[:, mask_idx, 30522:45473].squeeze(dim=0)
+        
+        grads = torch.autograd.grad(
+            # cross_entropy
+            torch.nn.functional.cross_entropy(
+                mask_logits[-1:, :],
+                torch.tensor([alter_label]),
+                reduction="none",
+            ).mean(-1),
+            ex_model.parameters(),
+        )
+
+    grads = {
+        name: grad
+        for (name, _), grad in zip(ex_model.named_parameters(), grads)
+    }
+
+    # cond_inputs里面有pad
+    params_dict = learner(
+        cond_inputs["input_ids"][-1:],
+        cond_inputs["attention_mask"][-1:],
+        grads=grads,
+    )
+
+    return params_dict
+
+def edit_process(edit_input, alter_label):
+    inputs, cond_inputs, edit_inputs = solve(edit_input, alter_label, edit_task=True)
+
+    _, mask_idx = (inputs["input_ids"] == tokenizer.mask_token_id).nonzero(as_tuple=True)
+    logits = edit_origin_model(**inputs).logits[:, :, 30522:45473].squeeze()
+    logits = logits[mask_idx, :]
+
+    ### origin output
+    _, origin_entity_order = torch.sort(logits, dim=1, descending=True)
+    origin_entity_order = origin_entity_order.squeeze(dim=0)
+    origin_top3 = [id2ent_name[id2ent[origin_entity_order[i].item()]] for i in range(3)]
+
+    ### edit output
+    fmodel = make_functional(edit_ex_model).eval()
+    params_dict = get_logits_orig_params_dict(inputs, cond_inputs, ent2id[ent_name2id[alter_label]], edit_ex_model, edit_learner)
+    edit_logits = fmodel(
+        input_ids=inputs["input_ids"],
+        attention_mask=inputs["attention_mask"],
+        # add delta theta
+        params=[
+            params_dict.get(n, 0) + p
+            for n, p in edit_ex_model.named_parameters()
+        ],
+    ).logits[:, :, 30522:45473].squeeze()
+    
+    edit_logits = edit_logits[mask_idx, :]
+    _, edit_entity_order = torch.sort(edit_logits, dim=1, descending=True)
+    edit_entity_order = edit_entity_order.squeeze(dim=0)
+    edit_top3 = [id2ent_name[id2ent[edit_entity_order[i].item()]] for i in range(3)]
+
+    return "\n".join(origin_top3), "\n".join(edit_top3)
+
+def add_process(edit_input, alter_label):
+    inputs, cond_inputs, add_inputs = solve(edit_input, alter_label, edit_task=False)
+
+    _, mask_idx = (inputs["input_ids"] == tokenizer.mask_token_id).nonzero(as_tuple=True)
+    logits = add_origin_model(**inputs).logits[:, :, 30522:45473].squeeze()
+    logits = logits[mask_idx, :]
+
+    ### origin output
+    _, origin_entity_order = torch.sort(logits, dim=1, descending=True)
+    origin_entity_order = origin_entity_order.squeeze(dim=0)
+    origin_top3 = [id2ent_name[id2ent[origin_entity_order[i].item()]] for i in range(3)]
+
+    ### add output
+    fmodel = make_functional(add_ex_model).eval()
+    params_dict = get_logits_orig_params_dict(inputs, cond_inputs, ent2id[ent_name2id[alter_label]], add_ex_model, add_learner)
+    add_logits = fmodel(
+        input_ids=inputs["input_ids"],
+        attention_mask=inputs["attention_mask"],
+        # add delta theta
+        params=[
+            params_dict.get(n, 0) + p
+            for n, p in add_ex_model.named_parameters()
+        ],
+    ).logits[:, :, 30522:45473].squeeze()
+    
+    add_logits = add_logits[mask_idx, :]
+    _, add_entity_order = torch.sort(add_logits, dim=1, descending=True)
+    add_entity_order = add_entity_order.squeeze(dim=0)
+    add_top3 = [id2ent_name[id2ent[add_entity_order[i].item()]] for i in range(3)]
+
+    return "\n".join(origin_top3), "\n".join(add_top3)
 
-def add_process(title, context, img):
-    return f"Title:{title}\nContext:{context}\n...{img}", f"Title:{title}\nContext:{context}\n...{img}"
 
 with gr.Blocks() as demo:
+    init_triple_input()
+    ### example
+    # edit_process("[MASK]|/people/person/profession|Jack Black", "Kellie Martin")
+    add_process("Red Skelton|/people/person/places_lived./people/place_lived/location|[MASK]", "Palm Springs")
     gr.Markdown("# KGE Editing")
 
     # 多个tab
     with gr.Tabs():
-
         with gr.TabItem("E-FB15k237"):
             with gr.Row():
                 with gr.Column():
@@ -25,7 +269,7 @@ with gr.Blocks() as demo:
                     edit_output = gr.Textbox(label="After Edit", lines=3, placeholder="")
        
             gr.Examples(
-                examples=[["[MASK] r1 t1", "h1"], ["[MASK] r2 t2", "h2"]],
+                examples=[["[MASK]|/people/person/profession|Jack Black", "Kellie Martin"], ["Red Skelton|/people/person/places_lived./people/place_lived/location|[MASK]", "Palm Springs"]],
                 inputs=[edit_input, alter_label],
                 outputs=[origin_output, edit_output],
                 fn=edit_process,
@@ -37,7 +281,7 @@ with gr.Blocks() as demo:
                 with gr.Column():
                     add_input = gr.Textbox(label="Input", lines=1, placeholder="New triple input")
 
-                    mask_head = gr.Textbox(label="Head/Tail", lines=1, placeholder="1:head / 0:tail")
+                    inductive_entity = gr.Textbox(label="Inductive Entity", lines=1, placeholder="Entity Name")
                     add_button = gr.Button("Add")
 
                 with gr.Column():
@@ -45,15 +289,14 @@ with gr.Blocks() as demo:
                     add_output = gr.Textbox(label="Add Results", lines=3, placeholder="")
 
             gr.Examples(
-                examples=[["h1 r1 t1", "1"], ["h2 r2 t2", "1"]],
-                inputs=[add_input, mask_head],
+                examples=[["Jane Wyman|/people/person/places_lived./people/place_lived/location|[MASK]", "Palm Springs"], ["Red Skelton|/people/person/places_lived./people/place_lived/location|[MASK]", "Palm Springs"]],
+                inputs=[add_input, inductive_entity],
                 outputs=[add_origin_output, add_output],
                 fn=add_process,
                 cache_examples=True,
             )
 
-    # origin_button.click(fn=origin_preditcion, inputs=[input, alter_label], outputs=origin_output)
     edit_button.click(fn=edit_process, inputs=[edit_input, alter_label], outputs=[origin_output, edit_output])
-    add_button.click(fn=add_process, inputs=[add_input, mask_head], outputs=[add_origin_output, add_output])
+    add_button.click(fn=add_process, inputs=[add_input, inductive_entity], outputs=[add_origin_output, add_output])
 
 demo.launch()

dataset/fb15k237/entity2textlong.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c6028d81296e311076eed7cfbf5dc3c4174b68394639148e32212ad49aa6c7f
+size 13063994

dataset/fb15k237/relations.txt

@@ -0,0 +1,237 @@
+/soccer/football_team/current_roster./soccer/football_roster_position/position
+/music/artist/origin
+/ice_hockey/hockey_team/current_roster./sports/sports_team_roster/position
+/food/food/nutrients./food/nutrition_fact/nutrient
+/film/actor/film./film/performance/film
+/award/award_nominee/award_nominations./award/award_nomination/nominated_for
+/government/political_party/politicians_in_this_party./government/political_party_tenure/politician
+/base/schemastaging/person_extra/net_worth./measurement_unit/dated_money_value/currency
+/people/deceased_person/place_of_death
+/people/person/profession
+/location/administrative_division/first_level_division_of
+/base/marchmadness/ncaa_basketball_tournament/seeds./base/marchmadness/ncaa_tournament_seed/team
+/education/university/international_tuition./measurement_unit/dated_money_value/currency
+/location/us_county/county_seat
+/location/location/partially_contains
+/tv/tv_program/program_creator
+/film/film/music
+/tv/tv_program/languages
+/common/topic/webpage./common/webpage/category
+/user/tsegaran/random/taxonomy_subject/entry./user/tsegaran/random/taxonomy_entry/taxonomy
+/education/field_of_study/students_majoring./education/education/major_field_of_study
+/business/business_operation/assets./measurement_unit/dated_money_value/currency
+/film/film_set_designer/film_sets_designed
+/dataworld/gardening_hint/split_to
+/people/person/languages
+/business/job_title/people_with_this_title./business/employment_tenure/company
+/location/country/form_of_government
+/base/schemastaging/organization_extra/phone_number./base/schemastaging/phone_sandbox/service_language
+/people/person/place_of_birth
+/sports/sports_team/colors
+/education/educational_institution/school_type
+/award/award_category/winners./award/award_honor/award_winner
+/organization/organization/headquarters./location/mailing_address/citytown
+/education/educational_degree/people_with_this_degree./education/education/student
+/government/legislative_session/members./government/government_position_held/legislative_sessions
+/film/film/distributors./film/film_film_distributor_relationship/film_distribution_medium
+/education/educational_degree/people_with_this_degree./education/education/major_field_of_study
+/location/hud_county_place/county
+/location/administrative_division/country
+/film/film/film_production_design_by
+/award/award_winning_work/awards_won./award/award_honor/award
+/organization/organization/headquarters./location/mailing_address/state_province_region
+/base/schemastaging/organization_extra/phone_number./base/schemastaging/phone_sandbox/contact_category
+/tv/tv_program/country_of_origin
+/olympics/olympic_participating_country/medals_won./olympics/olympic_medal_honor/medal
+/location/country/second_level_divisions
+/award/award_ceremony/awards_presented./award/award_honor/honored_for
+/organization/organization_member/member_of./organization/organization_membership/organization
+/education/educational_institution/campuses
+/music/artist/contribution./music/recording_contribution/performance_role
+/award/ranked_item/appears_in_ranked_lists./award/ranking/list
+/people/person/religion
+/travel/travel_destination/climate./travel/travel_destination_monthly_climate/month
+/film/special_film_performance_type/film_performance_type./film/performance/film
+/award/award_nominee/award_nominations./award/award_nomination/award
+/location/statistical_region/religions./location/religion_percentage/religion
+/sports/sports_league_draft/picks./sports/sports_league_draft_pick/school
+/film/film/distributors./film/film_film_distributor_relationship/region
+/government/politician/government_positions_held./government/government_position_held/legislative_sessions
+/organization/role/leaders./organization/leadership/organization
+/tv/tv_network/programs./tv/tv_network_duration/program
+/soccer/football_team/current_roster./sports/sports_team_roster/position
+/music/instrument/instrumentalists
+/business/business_operation/operating_income./measurement_unit/dated_money_value/currency
+/people/cause_of_death/people
+/film/film/film_art_direction_by
+/people/person/sibling_s./people/sibling_relationship/sibling
+/film/film/cinematography
+/film/actor/dubbing_performances./film/dubbing_performance/language
+/base/biblioness/bibs_location/state
+/base/petbreeds/city_with_dogs/top_breeds./base/petbreeds/dog_city_relationship/dog_breed
+/people/person/gender
+/education/field_of_study/students_majoring./education/education/student
+/base/popstra/celebrity/dated./base/popstra/dated/participant
+/sports/sports_team/roster./american_football/football_roster_position/position
+/award/award_winner/awards_won./award/award_honor/award_winner
+/olympics/olympic_participating_country/medals_won./olympics/olympic_medal_honor/olympics
+/film/director/film
+/tv/tv_producer/programs_produced./tv/tv_producer_term/program
+/film/film_distributor/films_distributed./film/film_film_distributor_relationship/film
+/olympics/olympic_games/sports
+/music/record_label/artist
+/education/university/local_tuition./measurement_unit/dated_money_value/currency
+/film/film/story_by
+/people/person/spouse_s./people/marriage/spouse
+/sports/sports_league/teams./sports/sports_league_participation/team
+/people/profession/specialization_of
+/base/americancomedy/celebrity_impressionist/celebrities_impersonated
+/tv/tv_program/genre
+/award/award_category/nominees./award/award_nomination/nominated_for
+/language/human_language/countries_spoken_in
+/organization/organization/headquarters./location/mailing_address/country
+/location/statistical_region/gdp_real./measurement_unit/adjusted_money_value/adjustment_currency
+/education/university/fraternities_and_sororities
+/award/award_nominee/award_nominations./award/award_nomination/award_nominee
+/military/military_combatant/military_conflicts./military/military_combatant_group/combatants
+/award/award_nominated_work/award_nominations./award/award_nomination/nominated_for
+/location/location/time_zones
+/film/film/dubbing_performances./film/dubbing_performance/actor
+/film/film_subject/films
+/education/educational_degree/people_with_this_degree./education/education/institution
+/education/educational_institution/colors
+/award/award_category/category_of
+/tv/tv_personality/tv_regular_appearances./tv/tv_regular_personal_appearance/program
+/film/film/language
+/music/group_member/membership./music/group_membership/group
+/business/business_operation/revenue./measurement_unit/dated_money_value/currency
+/film/film/film_festivals
+/film/actor/film./film/performance/special_performance_type
+/organization/non_profit_organization/registered_with./organization/non_profit_registration/registering_agency
+/government/politician/government_positions_held./government/government_position_held/jurisdiction_of_office
+/base/aareas/schema/administrative_area/administrative_parent
+/award/award_winning_work/awards_won./award/award_honor/award_winner
+/organization/organization/place_founded
+/soccer/football_player/current_team./sports/sports_team_roster/team
+/government/politician/government_positions_held./government/government_position_held/basic_title
+/music/artist/track_contributions./music/track_contribution/role
+/base/localfood/seasonal_month/produce_available./base/localfood/produce_availability/seasonal_months
+/celebrities/celebrity/celebrity_friends./celebrities/friendship/friend
+/sports/professional_sports_team/draft_picks./sports/sports_league_draft_pick/school
+/award/hall_of_fame/inductees./award/hall_of_fame_induction/inductee
+/influence/influence_node/peers./influence/peer_relationship/peers
+/medicine/disease/risk_factors
+/broadcast/content/artist
+/film/film/estimated_budget./measurement_unit/dated_money_value/currency
+/military/military_conflict/combatants./military/military_combatant_group/combatants
+/location/capital_of_administrative_division/capital_of./location/administrative_division_capital_relationship/administrative_division
+/tv/tv_program/regular_cast./tv/regular_tv_appearance/actor
+/people/deceased_person/place_of_burial
+/location/location/adjoin_s./location/adjoining_relationship/adjoins
+/music/group_member/membership./music/group_membership/role
+/award/award_ceremony/awards_presented./award/award_honor/award_winner
+/film/film/prequel
+/film/film/produced_by
+/tv/tv_program/tv_producer./tv/tv_producer_term/producer_type
+/sports/sports_position/players./sports/sports_team_roster/team
+/olympics/olympic_games/participating_countries
+/music/genre/parent_genre
+/tv/tv_writer/tv_programs./tv/tv_program_writer_relationship/tv_program
+/music/genre/artists
+/film/film/genre
+/people/person/employment_history./business/employment_tenure/company
+/education/university/domestic_tuition./measurement_unit/dated_money_value/currency
+/people/person/nationality
+/location/country/capital
+/location/statistical_region/gni_per_capita_in_ppp_dollars./measurement_unit/dated_money_value/currency
+/base/aareas/schema/administrative_area/capital
+/business/business_operation/industry
+/location/hud_foreclosure_area/estimated_number_of_mortgages./measurement_unit/dated_integer/source
+/film/film/other_crew./film/film_crew_gig/crewmember
+/base/popstra/location/vacationers./base/popstra/vacation_choice/vacationer
+/film/film/film_format
+/medicine/disease/notable_people_with_this_condition
+/film/film/costume_design_by
+/government/government_office_category/officeholders./government/government_position_held/jurisdiction_of_office
+/location/statistical_region/gdp_nominal./measurement_unit/dated_money_value/currency
+/sports/sports_team/roster./baseball/baseball_roster_position/position
+/award/award_winning_work/awards_won./award/award_honor/honored_for
+/olympics/olympic_sport/athletes./olympics/olympic_athlete_affiliation/olympics
+/celebrities/celebrity/sexual_relationships./celebrities/romantic_relationship/celebrity
+/people/marriage_union_type/unions_of_this_type./people/marriage/location_of_ceremony
+/organization/organization/child./organization/organization_relationship/child
+/organization/organization_founder/organizations_founded
+/sports/sports_team/sport
+/people/ethnicity/geographic_distribution
+/location/statistical_region/places_exported_to./location/imports_and_exports/exported_to
+/location/country/official_language
+/film/film/production_companies
+/user/jg/default_domain/olympic_games/sports
+/time/event/locations
+/people/person/spouse_s./people/marriage/type_of_union
+/government/governmental_body/members./government/government_position_held/legislative_sessions
+/media_common/netflix_genre/titles
+/user/alexander/philosophy/philosopher/interests
+/film/film/runtime./film/film_cut/film_release_region
+/education/educational_institution/students_graduates./education/education/student
+/base/eating/practicer_of_diet/diet
+/tv/non_character_role/tv_regular_personal_appearances./tv/tv_regular_personal_appearance/person
+/sports/sports_position/players./sports/sports_team_roster/position
+/sports/professional_sports_team/draft_picks./sports/sports_league_draft_pick/draft
+/medicine/symptom/symptom_of
+/film/person_or_entity_appearing_in_film/films./film/personal_film_appearance/type_of_appearance
+/sports/sports_team_location/teams
+/american_football/football_team/current_roster./sports/sports_team_roster/position
+/people/person/places_lived./people/place_lived/location
+/location/statistical_region/rent50_2./measurement_unit/dated_money_value/currency
+/film/film/personal_appearances./film/personal_film_appearance/person
+/music/instrument/family
+/sports/sports_team/roster./basketball/basketball_roster_position/position
+/base/schemastaging/organization_extra/phone_number./base/schemastaging/phone_sandbox/service_location
+/film/film/release_date_s./film/film_regional_release_date/film_release_region
+/award/award_category/disciplines_or_subjects
+/base/popstra/celebrity/friendship./base/popstra/friendship/participant
+/music/performance_role/regular_performances./music/group_membership/group
+/film/film/edited_by
+/base/x2010fifaworldcupsouthafrica/world_cup_squad/current_world_cup_squad./base/x2010fifaworldcupsouthafrica/current_world_cup_squad/current_club
+/base/popstra/celebrity/canoodled./base/popstra/canoodled/participant
+/film/film/release_date_s./film/film_regional_release_date/film_release_distribution_medium
+/film/film/other_crew./film/film_crew_gig/film_crew_role
+/base/popstra/celebrity/breakup./base/popstra/breakup/participant
+/film/film/country
+/music/performance_role/regular_performances./music/group_membership/role
+/sports/sports_team/roster./american_football/football_historical_roster_position/position_s
+/film/film/release_date_s./film/film_regional_release_date/film_regional_debut_venue
+/time/event/instance_of_recurring_event
+/olympics/olympic_participating_country/athletes./olympics/olympic_athlete_affiliation/olympics
+/organization/endowed_organization/endowment./measurement_unit/dated_money_value/currency
+/travel/travel_destination/how_to_get_here./travel/transportation/mode_of_transportation
+/baseball/baseball_team/team_stats./baseball/baseball_team_stats/season
+/award/award_category/winners./award/award_honor/ceremony
+/government/legislative_session/members./government/government_position_held/district_represented
+/influence/influence_node/influenced_by
+/base/culturalevent/event/entity_involved
+/people/ethnicity/people
+/sports/sport/pro_athletes./sports/pro_sports_played/athlete
+/location/statistical_region/gdp_nominal_per_capita./measurement_unit/dated_money_value/currency
+/location/hud_county_place/place
+/base/aareas/schema/administrative_area/administrative_area_type
+/base/locations/continents/countries_within
+/sports/sports_position/players./american_football/football_historical_roster_position/position_s
+/people/person/spouse_s./people/marriage/location_of_ceremony
+/education/educational_institution/students_graduates./education/education/major_field_of_study
+/film/film/written_by
+/olympics/olympic_sport/athletes./olympics/olympic_athlete_affiliation/country
+/music/performance_role/guest_performances./music/recording_contribution/performance_role
+/film/film/featured_film_locations
+/education/educational_institution_campus/educational_institution
+/sports/pro_athlete/teams./sports/sports_team_roster/team
+/people/ethnicity/languages_spoken
+/film/film/executive_produced_by
+/tv/tv_producer/programs_produced./tv/tv_producer_term/producer_type
+/location/location/contains
+/base/biblioness/bibs_location/country
+/user/ktrueman/default_domain/international_organization/member_states
+/music/performance_role/track_performances./music/track_contribution/role
+/olympics/olympic_games/medals_awarded./olympics/olympic_medal_honor/medal
+/base/saturdaynightlive/snl_cast_member/seasons./base/saturdaynightlive/snl_season_tenure/cast_members

requirement.txt

@@ -0,0 +1,5 @@
+allennlp
+kgeditor==1.0.0
+transformers
+jsonlines
+higher

src/modeling_bert.py

@@ -0,0 +1,1338 @@
+# coding=utf-8
+# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
+# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""PyTorch BERT model."""
+
+
+import math
+import os
+from dataclasses import dataclass
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.utils.checkpoint
+from packaging import version
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPastAndCrossAttentions,
+    BaseModelOutputWithPoolingAndCrossAttentions,
+    CausalLMOutputWithCrossAttentions,
+    MaskedLMOutput,
+    MultipleChoiceModelOutput,
+    NextSentencePredictorOutput,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutput,
+    TokenClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.pytorch_utils import apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer
+from transformers.utils import (
+    ModelOutput,
+    add_code_sample_docstrings,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    logging,
+)
+from transformers.models.bert.configuration_bert import BertConfig
+from transformers.activations import ACT2FN
+
+
+logger = logging.get_logger(__name__)
+
+_CHECKPOINT_FOR_DOC = "bert-base-uncased"
+_CONFIG_FOR_DOC = "BertConfig"
+_TOKENIZER_FOR_DOC = "BertTokenizer"
+
+# TokenClassification docstring
+_CHECKPOINT_FOR_TOKEN_CLASSIFICATION = "dbmdz/bert-large-cased-finetuned-conll03-english"
+_TOKEN_CLASS_EXPECTED_OUTPUT = (
+    "['O', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'O', 'O', 'O', 'O', 'I-LOC', 'O', 'I-LOC', 'I-LOC'] "
+)
+_TOKEN_CLASS_EXPECTED_LOSS = 0.01
+
+# QuestionAnswering docstring
+_CHECKPOINT_FOR_QA = "deepset/bert-base-cased-squad2"
+_QA_EXPECTED_OUTPUT = "'a nice puppet'"
+_QA_EXPECTED_LOSS = 7.41
+_QA_TARGET_START_INDEX = 14
+_QA_TARGET_END_INDEX = 15
+
+# SequenceClassification docstring
+_CHECKPOINT_FOR_SEQUENCE_CLASSIFICATION = "textattack/bert-base-uncased-yelp-polarity"
+_SEQ_CLASS_EXPECTED_OUTPUT = "'LABEL_1'"
+_SEQ_CLASS_EXPECTED_LOSS = 0.01
+
+
+BERT_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "bert-base-uncased",
+    "bert-large-uncased",
+    "bert-base-cased",
+    "bert-large-cased",
+    "bert-base-multilingual-uncased",
+    "bert-base-multilingual-cased",
+    "bert-base-chinese",
+    "bert-base-german-cased",
+    "bert-large-uncased-whole-word-masking",
+    "bert-large-cased-whole-word-masking",
+    "bert-large-uncased-whole-word-masking-finetuned-squad",
+    "bert-large-cased-whole-word-masking-finetuned-squad",
+    "bert-base-cased-finetuned-mrpc",
+    "bert-base-german-dbmdz-cased",
+    "bert-base-german-dbmdz-uncased",
+    "cl-tohoku/bert-base-japanese",
+    "cl-tohoku/bert-base-japanese-whole-word-masking",
+    "cl-tohoku/bert-base-japanese-char",
+    "cl-tohoku/bert-base-japanese-char-whole-word-masking",
+    "TurkuNLP/bert-base-finnish-cased-v1",
+    "TurkuNLP/bert-base-finnish-uncased-v1",
+    "wietsedv/bert-base-dutch-cased",
+    # See all BERT models at https://huggingface.co/models?filter=bert
+]
+
+
+def load_tf_weights_in_bert(model, config, tf_checkpoint_path):
+    """Load tf checkpoints in a pytorch model."""
+    try:
+        import re
+
+        import numpy as np
+        import tensorflow as tf
+    except ImportError:
+        logger.error(
+            "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
+            "https://www.tensorflow.org/install/ for installation instructions."
+        )
+        raise
+    tf_path = os.path.abspath(tf_checkpoint_path)
+    logger.info(f"Converting TensorFlow checkpoint from {tf_path}")
+    # Load weights from TF model
+    init_vars = tf.train.list_variables(tf_path)
+    names = []
+    arrays = []
+    for name, shape in init_vars:
+        logger.info(f"Loading TF weight {name} with shape {shape}")
+        array = tf.train.load_variable(tf_path, name)
+        names.append(name)
+        arrays.append(array)
+
+    for name, array in zip(names, arrays):
+        name = name.split("/")
+        # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
+        # which are not required for using pretrained model
+        if any(
+            n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"]
+            for n in name
+        ):
+            logger.info(f"Skipping {'/'.join(name)}")
+            continue
+        pointer = model
+        for m_name in name:
+            if re.fullmatch(r"[A-Za-z]+_\d+", m_name):
+                scope_names = re.split(r"_(\d+)", m_name)
+            else:
+                scope_names = [m_name]
+            if scope_names[0] == "kernel" or scope_names[0] == "gamma":
+                pointer = getattr(pointer, "weight")
+            elif scope_names[0] == "output_bias" or scope_names[0] == "beta":
+                pointer = getattr(pointer, "bias")
+            elif scope_names[0] == "output_weights":
+                pointer = getattr(pointer, "weight")
+            elif scope_names[0] == "squad":
+                pointer = getattr(pointer, "classifier")
+            else:
+                try:
+                    pointer = getattr(pointer, scope_names[0])
+                except AttributeError:
+                    logger.info(f"Skipping {'/'.join(name)}")
+                    continue
+            if len(scope_names) >= 2:
+                num = int(scope_names[1])
+                pointer = pointer[num]
+        if m_name[-11:] == "_embeddings":
+            pointer = getattr(pointer, "weight")
+        elif m_name == "kernel":
+            array = np.transpose(array)
+        try:
+            if pointer.shape != array.shape:
+                raise ValueError(f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched")
+        except AssertionError as e:
+            e.args += (pointer.shape, array.shape)
+            raise
+        logger.info(f"Initialize PyTorch weight {name}")
+        pointer.data = torch.from_numpy(array)
+    return model
+
+
+class BertEmbeddings(nn.Module):
+    """Construct the embeddings from word, position and token_type embeddings."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id)
+        self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size)
+        self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size)
+
+        # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load
+        # any TensorFlow checkpoint file
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        # position_ids (1, len position emb) is contiguous in memory and exported when serialized
+        self.position_embedding_type = getattr(config, "position_embedding_type", "absolute")
+        self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1)))
+        if version.parse(torch.__version__) > version.parse("1.6.0"):
+            self.register_buffer(
+                "token_type_ids",
+                torch.zeros(self.position_ids.size(), dtype=torch.long),
+                persistent=False,
+            )
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        token_type_ids: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        past_key_values_length: int = 0,
+    ) -> torch.Tensor:
+        if input_ids is not None:
+            input_shape = input_ids.size()
+        else:
+            input_shape = inputs_embeds.size()[:-1]
+
+        seq_length = input_shape[1]
+
+        if position_ids is None:
+            position_ids = self.position_ids[:, past_key_values_length : seq_length + past_key_values_length]
+
+        # Setting the token_type_ids to the registered buffer in constructor where it is all zeros, which usually occurs
+        # when its auto-generated, registered buffer helps users when tracing the model without passing token_type_ids, solves
+        # issue #5664
+        if token_type_ids is None:
+            if hasattr(self, "token_type_ids"):
+                buffered_token_type_ids = self.token_type_ids[:, :seq_length]
+                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(input_shape[0], seq_length)
+                token_type_ids = buffered_token_type_ids_expanded
+            else:
+                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+        token_type_embeddings = self.token_type_embeddings(token_type_ids)
+
+        embeddings = inputs_embeds + token_type_embeddings
+        if self.position_embedding_type == "absolute":
+            position_embeddings = self.position_embeddings(position_ids)
+            embeddings += position_embeddings
+        embeddings = self.LayerNorm(embeddings)
+        embeddings = self.dropout(embeddings)
+        return embeddings
+
+
+class BertSelfAttention(nn.Module):
+    def __init__(self, config, position_embedding_type=None):
+        super().__init__()
+        if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"):
+            raise ValueError(
+                f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention "
+                f"heads ({config.num_attention_heads})"
+            )
+
+        self.num_attention_heads = config.num_attention_heads
+        self.attention_head_size = int(config.hidden_size / config.num_attention_heads)
+        self.all_head_size = self.num_attention_heads * self.attention_head_size
+
+        self.query = nn.Linear(config.hidden_size, self.all_head_size)
+        self.key = nn.Linear(config.hidden_size, self.all_head_size)
+        self.value = nn.Linear(config.hidden_size, self.all_head_size)
+
+        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
+        self.position_embedding_type = position_embedding_type or getattr(
+            config, "position_embedding_type", "absolute"
+        )
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            self.max_position_embeddings = config.max_position_embeddings
+            self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size)
+
+        self.is_decoder = config.is_decoder
+
+    def transpose_for_scores(self, x: torch.Tensor) -> torch.Tensor:
+        new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size)
+        x = x.view(new_x_shape)
+        return x.permute(0, 2, 1, 3)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor]:
+        mixed_query_layer = self.query(hidden_states)
+
+        # If this is instantiated as a cross-attention module, the keys
+        # and values come from an encoder; the attention mask needs to be
+        # such that the encoder's padding tokens are not attended to.
+        is_cross_attention = encoder_hidden_states is not None
+
+        if is_cross_attention and past_key_value is not None:
+            # reuse k,v, cross_attentions
+            key_layer = past_key_value[0]
+            value_layer = past_key_value[1]
+            attention_mask = encoder_attention_mask
+        elif is_cross_attention:
+            key_layer = self.transpose_for_scores(self.key(encoder_hidden_states))
+            value_layer = self.transpose_for_scores(self.value(encoder_hidden_states))
+            attention_mask = encoder_attention_mask
+        elif past_key_value is not None:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+            key_layer = torch.cat([past_key_value[0], key_layer], dim=2)
+            value_layer = torch.cat([past_key_value[1], value_layer], dim=2)
+        else:
+            key_layer = self.transpose_for_scores(self.key(hidden_states))
+            value_layer = self.transpose_for_scores(self.value(hidden_states))
+
+        query_layer = self.transpose_for_scores(mixed_query_layer)
+
+        if self.is_decoder:
+            # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states.
+            # Further calls to cross_attention layer can then reuse all cross-attention
+            # key/value_states (first "if" case)
+            # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of
+            # all previous decoder key/value_states. Further calls to uni-directional self-attention
+            # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
+            # if encoder bi-directional self-attention `past_key_value` is always `None`
+            past_key_value = (key_layer, value_layer)
+
+        # Take the dot product between "query" and "key" to get the raw attention scores.
+        attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2))
+
+        if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query":
+            seq_length = hidden_states.size()[1]
+            position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1)
+            position_ids_r = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(1, -1)
+            distance = position_ids_l - position_ids_r
+            positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1)
+            positional_embedding = positional_embedding.to(dtype=query_layer.dtype)  # fp16 compatibility
+
+            if self.position_embedding_type == "relative_key":
+                relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores
+            elif self.position_embedding_type == "relative_key_query":
+                relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding)
+                relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding)
+                attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key
+
+        attention_scores = attention_scores / math.sqrt(self.attention_head_size)
+        if attention_mask is not None:
+            # Apply the attention mask is (precomputed for all layers in BertModel forward() function)
+            attention_scores = attention_scores + attention_mask
+
+        # Normalize the attention scores to probabilities.
+        attention_probs = nn.functional.softmax(attention_scores, dim=-1)
+
+        # This is actually dropping out entire tokens to attend to, which might
+        # seem a bit unusual, but is taken from the original Transformer paper.
+        attention_probs = self.dropout(attention_probs)
+
+        # Mask heads if we want to
+        if head_mask is not None:
+            attention_probs = attention_probs * head_mask
+
+        context_layer = torch.matmul(attention_probs, value_layer)
+
+        context_layer = context_layer.permute(0, 2, 1, 3).contiguous()
+        new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,)
+        context_layer = context_layer.view(new_context_layer_shape)
+
+        outputs = (context_layer, attention_probs) if output_attentions else (context_layer,)
+
+        if self.is_decoder:
+            outputs = outputs + (past_key_value,)
+        return outputs
+
+class BertSelfOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states + input_tensor)
+        return hidden_states
+
+class BertAttention(nn.Module):
+    def __init__(self, config, position_embedding_type=None):
+        super().__init__()
+        self.self = BertSelfAttention(config, position_embedding_type=position_embedding_type)
+        self.output = BertSelfOutput(config)
+        self.pruned_heads = set()
+
+    def prune_heads(self, heads):
+        if len(heads) == 0:
+            return
+        heads, index = find_pruneable_heads_and_indices(
+            heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads
+        )
+
+        # Prune linear layers
+        self.self.query = prune_linear_layer(self.self.query, index)
+        self.self.key = prune_linear_layer(self.self.key, index)
+        self.self.value = prune_linear_layer(self.self.value, index)
+        self.output.dense = prune_linear_layer(self.output.dense, index, dim=1)
+
+        # Update hyper params and store pruned heads
+        self.self.num_attention_heads = self.self.num_attention_heads - len(heads)
+        self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads
+        self.pruned_heads = self.pruned_heads.union(heads)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor]:
+        self_outputs = self.self(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            encoder_hidden_states,
+            encoder_attention_mask,
+            past_key_value,
+            output_attentions,
+        )
+        attention_output = self.output(self_outputs[0], hidden_states)
+        outputs = (attention_output,) + self_outputs[1:]  # add attentions if we output them
+        return outputs
+
+
+class BertIntermediate(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.intermediate_size)
+        if isinstance(config.hidden_act, str):
+            self.intermediate_act_fn = ACT2FN[config.hidden_act]
+        else:
+            self.intermediate_act_fn = config.hidden_act
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.intermediate_act_fn(hidden_states)
+        return hidden_states
+
+
+class BertOutputEx(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
+        
+        self.dense_in_ex = nn.Linear(config.intermediate_size, config.ex_size)
+        self.dense_out_ex = nn.Linear(config.ex_size, config.hidden_size)
+        
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
+        hidden_states_ori = self.dense(hidden_states)
+        
+        hidden_states_ex = self.dense_in_ex(hidden_states)
+        hidden_states_ex = self.dense_out_ex(hidden_states_ex)
+        
+        hidden_states_ori = self.dropout(hidden_states_ori)
+        
+        hidden_states = self.LayerNorm(hidden_states_ori + hidden_states_ex + input_tensor)
+        return hidden_states
+
+class BertOutput(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.intermediate_size, config.hidden_size)
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+
+    def forward(self, hidden_states: torch.Tensor, input_tensor: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states + input_tensor)
+        return hidden_states
+
+class BertLayerEx(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = BertAttention(config)
+        self.is_decoder = config.is_decoder
+        self.add_cross_attention = config.add_cross_attention
+        if self.add_cross_attention:
+            if not self.is_decoder:
+                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
+            self.crossattention = BertAttention(config, position_embedding_type="absolute")
+        self.intermediate = BertIntermediate(config)
+        self.output = BertOutputEx(config)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor]:
+        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
+        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            output_attentions=output_attentions,
+            past_key_value=self_attn_past_key_value,
+        )
+        attention_output = self_attention_outputs[0]
+
+        # if decoder, the last output is tuple of self-attn cache
+        if self.is_decoder:
+            outputs = self_attention_outputs[1:-1]
+            present_key_value = self_attention_outputs[-1]
+        else:
+            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        cross_attn_present_key_value = None
+        if self.is_decoder and encoder_hidden_states is not None:
+            if not hasattr(self, "crossattention"):
+                raise ValueError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
+                    " by setting `config.add_cross_attention=True`"
+                )
+
+            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
+            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
+            cross_attention_outputs = self.crossattention(
+                attention_output,
+                attention_mask,
+                head_mask,
+                encoder_hidden_states,
+                encoder_attention_mask,
+                cross_attn_past_key_value,
+                output_attentions,
+            )
+            attention_output = cross_attention_outputs[0]
+            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights
+
+            # add cross-attn cache to positions 3,4 of present_key_value tuple
+            cross_attn_present_key_value = cross_attention_outputs[-1]
+            present_key_value = present_key_value + cross_attn_present_key_value
+
+        layer_output = apply_chunking_to_forward(
+            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
+        )
+        outputs = (layer_output,) + outputs
+
+        # if decoder, return the attn key/values as the last output
+        if self.is_decoder:
+            outputs = outputs + (present_key_value,)
+
+        return outputs
+
+    def feed_forward_chunk(self, attention_output):
+        intermediate_output = self.intermediate(attention_output)
+        layer_output = self.output(intermediate_output, attention_output)
+        return layer_output
+
+class BertLayer(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.chunk_size_feed_forward = config.chunk_size_feed_forward
+        self.seq_len_dim = 1
+        self.attention = BertAttention(config)
+        self.is_decoder = config.is_decoder
+        self.add_cross_attention = config.add_cross_attention
+        if self.add_cross_attention:
+            if not self.is_decoder:
+                raise ValueError(f"{self} should be used as a decoder model if cross attention is added")
+            self.crossattention = BertAttention(config, position_embedding_type="absolute")
+        self.intermediate = BertIntermediate(config)
+        self.output = BertOutput(config)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_value: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor]:
+        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
+        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
+        self_attention_outputs = self.attention(
+            hidden_states,
+            attention_mask,
+            head_mask,
+            output_attentions=output_attentions,
+            past_key_value=self_attn_past_key_value,
+        )
+        attention_output = self_attention_outputs[0]
+
+        # if decoder, the last output is tuple of self-attn cache
+        if self.is_decoder:
+            outputs = self_attention_outputs[1:-1]
+            present_key_value = self_attention_outputs[-1]
+        else:
+            outputs = self_attention_outputs[1:]  # add self attentions if we output attention weights
+
+        cross_attn_present_key_value = None
+        if self.is_decoder and encoder_hidden_states is not None:
+            if not hasattr(self, "crossattention"):
+                raise ValueError(
+                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with cross-attention layers"
+                    " by setting `config.add_cross_attention=True`"
+                )
+
+            # cross_attn cached key/values tuple is at positions 3,4 of past_key_value tuple
+            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
+            cross_attention_outputs = self.crossattention(
+                attention_output,
+                attention_mask,
+                head_mask,
+                encoder_hidden_states,
+                encoder_attention_mask,
+                cross_attn_past_key_value,
+                output_attentions,
+            )
+            attention_output = cross_attention_outputs[0]
+            outputs = outputs + cross_attention_outputs[1:-1]  # add cross attentions if we output attention weights
+
+            # add cross-attn cache to positions 3,4 of present_key_value tuple
+            cross_attn_present_key_value = cross_attention_outputs[-1]
+            present_key_value = present_key_value + cross_attn_present_key_value
+
+        layer_output = apply_chunking_to_forward(
+            self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output
+        )
+        outputs = (layer_output,) + outputs
+
+        # if decoder, return the attn key/values as the last output
+        if self.is_decoder:
+            outputs = outputs + (present_key_value,)
+
+        return outputs
+
+    def feed_forward_chunk(self, attention_output):
+        intermediate_output = self.intermediate(attention_output)
+        layer_output = self.output(intermediate_output, attention_output)
+        return layer_output
+
+class BertEncoder(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        kb_layer = self.config.kb_layer
+        self.layer = nn.ModuleList([BertLayerEx(config) if i in kb_layer else BertLayer(config) for i in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.FloatTensor] = None,
+        head_mask: Optional[torch.FloatTensor] = None,
+        encoder_hidden_states: Optional[torch.FloatTensor] = None,
+        encoder_attention_mask: Optional[torch.FloatTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = False,
+        output_hidden_states: Optional[bool] = False,
+        return_dict: Optional[bool] = True,
+    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPastAndCrossAttentions]:
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attentions = () if output_attentions else None
+        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
+
+        next_decoder_cache = () if use_cache else None
+        for i, layer_module in enumerate(self.layer):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            layer_head_mask = head_mask[i] if head_mask is not None else None
+            past_key_value = past_key_values[i] if past_key_values is not None else None
+
+            if self.gradient_checkpointing and self.training:
+
+                if use_cache:
+                    logger.warning(
+                        "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                    )
+                    use_cache = False
+
+                def create_custom_forward(module):
+                    def custom_forward(*inputs):
+                        return module(*inputs, past_key_value, output_attentions)
+
+                    return custom_forward
+
+                layer_outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(layer_module),
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                )
+            else:
+                layer_outputs = layer_module(
+                    hidden_states,
+                    attention_mask,
+                    layer_head_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    past_key_value,
+                    output_attentions,
+                )
+
+            hidden_states = layer_outputs[0]
+            if use_cache:
+                next_decoder_cache += (layer_outputs[-1],)
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (layer_outputs[1],)
+                if self.config.add_cross_attention:
+                    all_cross_attentions = all_cross_attentions + (layer_outputs[2],)
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [
+                    hidden_states,
+                    next_decoder_cache,
+                    all_hidden_states,
+                    all_self_attentions,
+                    all_cross_attentions,
+                ]
+                if v is not None
+            )
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=next_decoder_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+            cross_attentions=all_cross_attentions,
+        )
+
+
+class BertPooler(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.activation = nn.Tanh()
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # We "pool" the model by simply taking the hidden state corresponding
+        # to the first token.
+        first_token_tensor = hidden_states[:, 0]
+        pooled_output = self.dense(first_token_tensor)
+        pooled_output = self.activation(pooled_output)
+        return pooled_output
+
+class BertPredictionHeadTransform(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        if isinstance(config.hidden_act, str):
+            self.transform_act_fn = ACT2FN[config.hidden_act]
+        else:
+            self.transform_act_fn = config.hidden_act
+        self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.dense(hidden_states)
+        hidden_states = self.transform_act_fn(hidden_states)
+        hidden_states = self.LayerNorm(hidden_states)
+        return hidden_states
+
+class BertLMPredictionHead(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.transform = BertPredictionHeadTransform(config)
+
+        # The output weights are the same as the input embeddings, but there is
+        # an output-only bias for each token.
+        self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        self.bias = nn.Parameter(torch.zeros(config.vocab_size))
+
+        # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings`
+        self.decoder.bias = self.bias
+
+    def forward(self, hidden_states):
+        hidden_states = self.transform(hidden_states)
+        hidden_states = self.decoder(hidden_states)
+        return hidden_states
+
+class BertOnlyMLMHead(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.predictions = BertLMPredictionHead(config)
+
+    def forward(self, sequence_output: torch.Tensor) -> torch.Tensor:
+        prediction_scores = self.predictions(sequence_output)
+        return prediction_scores
+      
+class BertPreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = BertConfig
+    load_tf_weights = load_tf_weights_in_bert
+    base_model_prefix = "bert"
+    supports_gradient_checkpointing = True
+    _keys_to_ignore_on_load_missing = [r"position_ids"]
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            # Slightly different from the TF version which uses truncated_normal for initialization
+            # cf https://github.com/pytorch/pytorch/pull/5617
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if isinstance(module, BertEncoder):
+            module.gradient_checkpointing = value
+
+
+@dataclass
+class BertForPreTrainingOutput(ModelOutput):
+    """
+    Output type of [`BertForPreTraining`].
+
+    Args:
+        loss (*optional*, returned when `labels` is provided, `torch.FloatTensor` of shape `(1,)`):
+            Total loss as the sum of the masked language modeling loss and the next sequence prediction
+            (classification) loss.
+        prediction_logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
+            Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+        seq_relationship_logits (`torch.FloatTensor` of shape `(batch_size, 2)`):
+            Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation
+            before SoftMax).
+        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) of
+            shape `(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    prediction_logits: torch.FloatTensor = None
+    seq_relationship_logits: torch.FloatTensor = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+BERT_START_DOCSTRING = r"""
+
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`BertConfig`]): Model configuration class with all the parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+BERT_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `({0})`):
+            Indices of input sequence tokens in the vocabulary.
+
+            Indices can be obtained using [`BertTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.FloatTensor` of shape `({0})`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+        token_type_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+            Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
+            1]`:
+
+            - 0 corresponds to a *sentence A* token,
+            - 1 corresponds to a *sentence B* token.
+
+            [What are token type IDs?](../glossary#token-type-ids)
+        position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.max_position_embeddings - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
+            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+
+        inputs_embeds (`torch.FloatTensor` of shape `({0}, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+@add_start_docstrings(
+    "The bare Bert Model transformer outputting raw hidden-states without any specific head on top.",
+    BERT_START_DOCSTRING,
+)
+class BertModel(BertPreTrainedModel):
+    """
+
+    The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of
+    cross-attention is added between the self-attention layers, following the architecture described in [Attention is
+    all you need](https://arxiv.org/abs/1706.03762) by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit,
+    Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.
+
+    To behave as an decoder the model needs to be initialized with the `is_decoder` argument of the configuration set
+    to `True`. To be used in a Seq2Seq model, the model needs to initialized with both `is_decoder` argument and
+    `add_cross_attention` set to `True`; an `encoder_hidden_states` is then expected as an input to the forward pass.
+    """
+
+    def __init__(self, config, add_pooling_layer=True):
+        super().__init__(config)
+        self.config = config
+
+        self.embeddings = BertEmbeddings(config)
+        self.encoder = BertEncoder(config)
+
+        self.pooler = BertPooler(config) if add_pooling_layer else None
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embeddings.word_embeddings
+
+    def set_input_embeddings(self, value):
+        self.embeddings.word_embeddings = value
+
+    def _prune_heads(self, heads_to_prune):
+        """
+        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base
+        class PreTrainedModel
+        """
+        for layer, heads in heads_to_prune.items():
+            self.encoder.layer[layer].attention.prune_heads(heads)
+
+    @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        processor_class=_TOKENIZER_FOR_DOC,
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=BaseModelOutputWithPoolingAndCrossAttentions,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAttentions]:
+        r"""
+        encoder_hidden_states  (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
+            the model is configured as a decoder.
+        encoder_attention_mask (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
+            the cross-attention if the model is configured as a decoder. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+        past_key_values (`tuple(tuple(torch.FloatTensor))` of length `config.n_layers` with each tuple having 4 tensors of shape `(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
+            Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
+
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if self.config.is_decoder:
+            use_cache = use_cache if use_cache is not None else self.config.use_cache
+        else:
+            use_cache = False
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            input_shape = input_ids.size()
+        elif inputs_embeds is not None:
+            input_shape = inputs_embeds.size()[:-1]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        batch_size, seq_length = input_shape
+        device = input_ids.device if input_ids is not None else inputs_embeds.device
+
+        # past_key_values_length
+        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
+
+        if attention_mask is None:
+            attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device)
+
+        if token_type_ids is None:
+            if hasattr(self.embeddings, "token_type_ids"):
+                buffered_token_type_ids = self.embeddings.token_type_ids[:, :seq_length]
+                buffered_token_type_ids_expanded = buffered_token_type_ids.expand(batch_size, seq_length)
+                token_type_ids = buffered_token_type_ids_expanded
+            else:
+                token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
+
+        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
+        # ourselves in which case we just need to make it broadcastable to all heads.
+        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape)
+
+        # If a 2D or 3D attention mask is provided for the cross-attention
+        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
+        if self.config.is_decoder and encoder_hidden_states is not None:
+            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
+            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
+            if encoder_attention_mask is None:
+                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
+            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
+        else:
+            encoder_extended_attention_mask = None
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape bsz x n_heads x N x N
+        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
+        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
+        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
+
+        embedding_output = self.embeddings(
+            input_ids=input_ids,
+            position_ids=position_ids,
+            token_type_ids=token_type_ids,
+            inputs_embeds=inputs_embeds,
+            past_key_values_length=past_key_values_length,
+        )
+        encoder_outputs = self.encoder(
+            embedding_output,
+            attention_mask=extended_attention_mask,
+            head_mask=head_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_extended_attention_mask,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        sequence_output = encoder_outputs[0]
+        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
+
+        if not return_dict:
+            return (sequence_output, pooled_output) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPoolingAndCrossAttentions(
+            last_hidden_state=sequence_output,
+            pooler_output=pooled_output,
+            past_key_values=encoder_outputs.past_key_values,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+            cross_attentions=encoder_outputs.cross_attentions,
+        )
+
+@add_start_docstrings("""Bert Model with a `language modeling` head on top.""", BERT_START_DOCSTRING)
+class EXBertForMaskedLM(BertPreTrainedModel):
+
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+    _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"]
+
+    def __init__(self, config):
+        super().__init__(config)
+
+        if config.is_decoder:
+            logger.warning(
+                "If you want to use `BertForMaskedLM` make sure `config.is_decoder=False` for "
+                "bi-directional self-attention."
+            )
+
+        self.bert = BertModel(config, add_pooling_layer=False)
+        self.cls = BertOnlyMLMHead(config)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_output_embeddings(self):
+        return self.cls.predictions.decoder
+
+    def set_output_embeddings(self, new_embeddings):
+        self.cls.predictions.decoder = new_embeddings
+
+    @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        processor_class=_TOKENIZER_FOR_DOC,
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=MaskedLMOutput,
+        config_class=_CONFIG_FOR_DOC,
+        expected_output="'paris'",
+        expected_loss=0.88,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], MaskedLMOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
+            config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
+            loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
+        """
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+        prediction_scores = self.cls(sequence_output) 
+        # pos = (input_ids == self.config.mask_token_id).nonzero(as_tuple=True)
+
+        masked_lm_loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()  # -100 index = padding token
+            masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
+            # masked_lm_loss = loss_fct(prediction_scores[pos[0], pos[1], :].view(-1, self.config.vocab_size), labels.view(-1))
+
+        if not return_dict:
+            output = (prediction_scores,) + outputs[2:]
+            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output
+
+        return MaskedLMOutput(
+            loss=masked_lm_loss,
+            logits=prediction_scores,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **model_kwargs):
+        input_shape = input_ids.shape
+        effective_batch_size = input_shape[0]
+
+        #  add a dummy token
+        if self.config.pad_token_id is None:
+            raise ValueError("The PAD token should be defined for generation")
+
+        attention_mask = torch.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], dim=-1)
+        dummy_token = torch.full(
+            (effective_batch_size, 1), self.config.pad_token_id, dtype=torch.long, device=input_ids.device
+        )
+        input_ids = torch.cat([input_ids, dummy_token], dim=1)
+
+        return {"input_ids": input_ids, "attention_mask": attention_mask}
+
+@add_start_docstrings(
+    """
+    Bert Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled
+    output) e.g. for GLUE tasks.
+    """,
+    BERT_START_DOCSTRING,
+)
+class EXBertForSequenceClassification(BertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.config = config
+
+        self.bert = BertModel(config)
+        classifier_dropout = (
+            config.classifier_dropout if config.classifier_dropout is not None else config.hidden_dropout_prob
+        )
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @add_code_sample_docstrings(
+        processor_class=_TOKENIZER_FOR_DOC,
+        checkpoint=_CHECKPOINT_FOR_SEQUENCE_CLASSIFICATION,
+        output_type=SequenceClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+        expected_output=_SEQ_CLASS_EXPECTED_OUTPUT,
+        expected_loss=_SEQ_CLASS_EXPECTED_LOSS,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], SequenceClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        pooled_output = outputs[1]
+
+        pooled_output = self.dropout(pooled_output)
+        logits = self.classifier(pooled_output)
+
+        loss = None
+        if labels is not None:
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(logits, labels)
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+bert_mapping = {
+  'FT': EXBertForSequenceClassification,
+  'PT': EXBertForMaskedLM
+}

src/models/one_shot_learner.py

@@ -0,0 +1,157 @@
+import torch
+from allennlp.modules.feedforward import FeedForward
+from allennlp.modules.seq2vec_encoders import PytorchSeq2VecWrapper
+from higher.patch import monkeypatch as make_functional
+
+
+class ConditionedParameter(torch.nn.Module):
+    def __init__(self, parameter, condition_dim=1024, hidden_dim=128, max_scale=1):
+        super().__init__()
+        self.parameter_shape = parameter.shape
+
+        if len(self.parameter_shape) == 2: # condition_dim是从lstm中得到的tensor，然后用linear学习返回到768作为更新的parm_dict
+            self.conditioners = torch.nn.Sequential(
+                torch.nn.utils.weight_norm(torch.nn.Linear(condition_dim, hidden_dim)),
+                torch.nn.Tanh(),
+                torch.nn.utils.weight_norm(
+                    torch.nn.Linear(
+                        hidden_dim, 2 * (parameter.shape[0] + parameter.shape[1]) + 1
+                    )
+                ),
+            )
+        elif len(self.parameter_shape) == 1:
+            self.conditioners = torch.nn.Sequential(
+                torch.nn.utils.weight_norm(torch.nn.Linear(condition_dim, hidden_dim)),
+                torch.nn.Tanh(),
+                torch.nn.utils.weight_norm(
+                    torch.nn.Linear(hidden_dim, 2 * parameter.shape[0] + 1)
+                ),
+            )
+        else:
+            raise RuntimeError()
+
+        self.max_scale = max_scale
+
+    def forward(self, inputs, grad):
+
+        if len(self.parameter_shape) == 2:
+            (
+                conditioner_cola,
+                conditioner_rowa,
+                conditioner_colb,
+                conditioner_rowb,
+                conditioner_norm,
+            ) = self.conditioners(inputs).split(
+                [
+                    self.parameter_shape[1],
+                    self.parameter_shape[0],
+                    self.parameter_shape[1],
+                    self.parameter_shape[0],
+                    1,
+                ],
+                dim=-1,
+            )
+
+            a = conditioner_rowa.softmax(-1).T @ conditioner_cola
+            b = conditioner_rowb.softmax(-1).T @ conditioner_colb
+
+        elif len(self.parameter_shape) == 1:
+            a, b, conditioner_norm = self.conditioners(inputs).split(
+                [self.parameter_shape[0], self.parameter_shape[0], 1], dim=-1
+            )
+        else:
+            raise RuntimeError()
+
+        return (
+            self.max_scale
+            * torch.mean(conditioner_norm.sigmoid(), dim=0).squeeze() # 多条我们直接取mean
+            * (grad * a.squeeze() + b.squeeze())
+        )
+
+
+class LSTMConditioner(torch.nn.Module):
+    def __init__(
+        self,
+        vocab_dim=30522,
+        embedding_dim=768,
+        hidden_dim=256,
+        output_dim=1024,
+        embedding_init=None,
+    ):
+        super().__init__()
+        self.embedding = torch.nn.Embedding(
+            num_embeddings=vocab_dim,
+            embedding_dim=embedding_dim,
+            padding_idx=0,
+            _weight=embedding_init,
+        )
+        self.lstm = PytorchSeq2VecWrapper(
+            torch.nn.LSTM(
+                input_size=embedding_dim,
+                hidden_size=hidden_dim,
+                num_layers=1,
+                bidirectional=True,
+                batch_first=True,
+            )
+        )
+        self.linear = FeedForward(
+            input_dim=hidden_dim * 2,
+            num_layers=1,
+            hidden_dims=[output_dim],
+            activations=[torch.nn.Tanh()],
+        )
+
+    def forward(self, inputs, masks):
+        return self.linear(self.lstm(self.embedding(inputs), masks)) # 1, 64
+
+
+class OneShotLearner(torch.nn.Module):
+    def __init__(
+        self,
+        model,
+        vocab_dim=30522,
+        embedding_dim=768,
+        hidden_dim=128,
+        condition_dim=1024,
+        include_set={},
+        max_scale=1e-3,
+        embedding_init=None,
+    ):
+        super().__init__()
+
+        self.param2conditioner_map = {
+            n: "{}_conditioner".format(n).replace(".", "_")
+            for n, p in model.named_parameters()
+            if n in include_set
+        }
+
+        self.conditioners = torch.nn.ModuleDict(
+            {
+                self.param2conditioner_map[n]: ConditionedParameter(
+                    p,
+                    condition_dim,
+                    hidden_dim,
+                    max_scale=max_scale,
+                )
+                for n, p in model.named_parameters()
+                if n in include_set
+            }
+        )
+
+        self.condition = LSTMConditioner(
+            vocab_dim,
+            embedding_dim,
+            hidden_dim,
+            condition_dim,
+            embedding_init=embedding_init,
+        )
+
+    def forward(self, inputs, masks, grads=None):
+        condition = self.condition(inputs, masks) # LSTM输出condition
+        return {
+            p: self.conditioners[self.param2conditioner_map[p]](
+                condition,
+                grad=grads[p] if grads else None,
+            )
+            for p, c in self.param2conditioner_map.items()
+        }