tiny_bert_ft_sim_score_1231 / README.md

Tien09

Add new SentenceTransformer model

3f53316 verified about 1 year ago

preview code

raw

history blame contribute delete

32.1 kB

metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:8959
  - loss:CoSENTLoss
base_model: prajjwal1/bert-tiny
widget:
  - source_sentence: >-
      This card is treated as a Normal Monster while face-up on the field or in
      the GY. While this card is face-up on the field, you can Normal Summon it
      to have it become an Effect Monster with this effect. During your End
      Phase: You can target 1 Equip Spell Card in your GY; add that target to
      your hand. You can only use this effect of "Knight Day Grepher" once per
      turn.
    sentences:
      - >-
        Cannot be Normal Summoned/Set. Must first be Special Summoned (from your
        hand) while the total ATK of all monsters your opponent controls is more
        than that of all monsters you control. You can target any number of
        Beast, Beast-Warrior, and/or Winged Beast monsters you control; return
        them to the hand, then, return face-up monsters your opponent controls
        to the hand, equal to the number of cards you returned to the hand, also
        for the rest of this turn, "Alpha, the Master of Beasts" you control
        cannot attack directly. You can only use this effect of "Alpha, the
        Master of Beasts" once per turn.
      - "(This card is always treated as an \"Infernoble Knight\" card.)\r\nYou can discard this card, then target 1 Warrior monster you control; equip it with 1 Equip Spell from your Deck that can equip to it. When this card, not equipped with an Equip Card, is destroyed by battle with an opponent's monster and sent to the GY: You can Special Summon this card, and if you do, equip that monster your opponent controls to this card. You can only use each effect of \"Courageous Crimson Chevalier Bradamante\" once per turn."
      - >-
        If you have no cards in your GY (Quick Effect): You can send this card
        from your hand to the GY; until the end of the next turn, any card sent
        to the GY is banished instead.
  - source_sentence: >-
      2 Cyberse monsters If this card is Link Summoned: You can add 1 "Cynet
      Fusion" from your Deck to your hand. If a monster(s) is Special Summoned
      to a zone(s) this card points to (except during the Damage Step): You can
      target 1 Level 4 or lower Cyberse monster in your GY; Special Summon it,
      but negate its effects, also you cannot Special Summon monsters from the
      Extra Deck for the rest of this turn, except Fusion Monsters. You can only
      use each effect of "Clock Spartoi" once per turn.
    sentences:
      - >-
        This card cannot be Normal Summoned or Set. This card cannot be Special
        Summoned except by its own effect. When a face-up monster you control is
        destroyed by a card effect and sent to the GY, you can Special Summon
        this card from your hand. This card gains ATK and DEF equal to half of
        your Life Points. Once per turn, you can select 1 Synchro Monster your
        opponent controls and equip it to this card. This card gains ATK equal
        to those equipped monsters' combined ATK. During your Main Phase, you
        can select 1 of those equipped monsters, and Special Summon it to your
        side of the field in face-up Defense Position.
      - >-
        If this card is Special Summoned: You can discard 1 card, and if you do,
        send 1 "Fossil" Fusion Monster from your Extra Deck to the GY. You can
        send this card you control to the GY; inflict 500 damage to your
        opponent, then, if "Fossil Fusion" is in your GY, you can return 1 of
        your banished "Fossil Fusion", or 1 card that specifically lists that
        card in its text, to the GY. You can only use each effect of "Flint
        Cragger" once per turn.
      - >-
        If this card is sent from the Spell & Trap Zone to the GY: You can
        target 1 monster on the field; destroy it. You cannot activate the
        following effects of "True Draco Apocalypse" in the same Chain.

        ● You can target 1 other "True Draco" or "True King" card you control;
        destroy it, and if you do, the ATK and DEF of all face-up monsters your
        opponent controls become half their current ATK and DEF (even if this
        card leaves the field).

        ● During your opponent's Main Phase, you can: Immediately after this
        effect resolves, Tribute Summon 1 "True Draco" or "True King" monster
        face-up. You can only use each effect of "True Draco Apocalypse" once
        per turn.
  - source_sentence: >-
      When this card is Normal Summoned, place 1 Spell counter on it (max. 1).
      You can remove 1 Spell counter from this card to have 1 face-up monster
      you control gain 500 ATK, until the End Phase.
    sentences:
      - >-
        Place 1 Ocean Counter on this card during each player's Standby Phases.
        When this card is removed from the field, all Fish-Type and Sea
        Serpent-Type monsters you control gain 200 ATK for each Ocean Counter on
        this card, until the End Phase.
      - "Apply this effect, then you can discard 1 card, and if you do, Special Summon 1 Level 1 \"Purrely\" monster from your Deck. Once applied, the next battle or effect damage you take this turn becomes 0. A \"Purrely\" Xyz Monster that has this card as material gains this effect.\r\n● Once per turn, during your opponent's Standby Phase: You can draw 1 card."
      - >-
        Choose 1 Fusion Monster you control that mentions "Fallen of Albaz" as
        material, and negate the effects of all other face-up cards currently on
        the field until the end of this turn. During the End Phase, if a Fusion
        Monster(s) was sent to your GY this turn: You can add this card from the
        GY to your hand. You can only use 1 "Brightest, Blazing, Branded King"
        effect per turn, and only once that turn.
  - source_sentence: >-
      When an attack is declared involving your face-up monster and an
      opponent's monster: You can Special Summon this card from your hand or GY,
      but banish it when it leaves the field. You can only use this effect of
      "Vic Viper T301" once per turn. Other Light Machine monsters you control
      gain 1200 ATK.
    sentences:
      - "Banish 1 \"Springans\", \"Therion\" or \"Spright\" card from your hand or GY, then activate 1 of these effects;\r\n● Special Summon 1 \"Springans\" monster from your Deck.\r\n● Special Summon 1 \"Therion\" monster from your GY.\r\n● Banish 1 Level/Rank/Link 2 monster you control and 1 card your opponent controls.\r\nYou can only activate 1 \"Spright Smashers\" per turn."
      - When this card is discarded from your hand to the GY, Special Summon it.
      - >-
        When this card is activated: You can place 1 or 2 "Dinomist" Pendulum
        Monster(s) from your Deck in your Pendulum Zone, but you cannot Pendulum
        Summon monsters until the end of the next turn, except "Dinomist"
        monsters (even if this card leaves the field). Once per turn, if this
        card is already face-up in your Spell & Trap Zone: You can Tribute 1
        "Dinomist" monster, then target 1 card your opponent controls; return it
        to the hand.
  - source_sentence: >-
      You can target 1 face-up card you control; destroy it, and if you do,
      Special Summon 1 "Zoodiac" monster from your Deck. You can only use this
      effect of "Zoodiac Barrage" once per turn. If this card is destroyed by a
      card effect and sent to the GY: You can target 1 "Zoodiac" Xyz Monster you
      control; attach this card from your GY to that Xyz Monster as Xyz
      Material.
    sentences:
      - >-
        Destroy as many Normal Monsters on the field as possible, and if you do,
        Special Summon Level 4 or lower Dinosaur-Type monsters from your Deck,
        up to the number destroyed, but destroy them during the End Phase. You
        can banish this card from your GY, then target 1 Dinosaur-Type monster
        you control and 1 card your opponent controls; destroy them.
      - "2 monsters with different Types\r\nDuring the Main Phase (Quick Effect): You can Tribute this card and 1 Plant monster; Special Summon 1 \"Rose\" Synchro Monster or 1 Plant Synchro Monster from your Extra Deck. (This is treated as a Synchro Summon.) If a monster(s) you control is destroyed by card effect, while this card is in your GY (except during the Damage Step): You can banish this card; Special Summon 1 \"Rose Dragon\" monster from your GY. You can only use each effect of \"Crossrose Dragon\" once per turn."
      - >-
        Target 1 "Raidraptor" monster you control; Special Summon 1 monster with
        the same name as that monster on the field from your hand or Deck in
        Defense Position. You can only activate 1 "Raidraptor - Call" per turn.
        You cannot Special Summon monsters during the turn you activate this
        card, except "Raidraptor" monsters.
datasets:
  - Tien09/pair_similarity_new_1231
pipeline_tag: sentence-similarity
library_name: sentence-transformers

MPNet base trained on AllNLI triplets

This is a sentence-transformers model finetuned from prajjwal1/bert-tiny on the pair_similarity_new_1231 dataset. It maps sentences & paragraphs to a 128-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: prajjwal1/bert-tiny
Maximum Sequence Length: 512 tokens
Output Dimensionality: 128 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- pair_similarity_new_1231
Language: en
License: apache-2.0

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 128, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Tien09/tiny_bert_ft_sim_score_1231")
# Run inference
sentences = [
    'You can target 1 face-up card you control; destroy it, and if you do, Special Summon 1 "Zoodiac" monster from your Deck. You can only use this effect of "Zoodiac Barrage" once per turn. If this card is destroyed by a card effect and sent to the GY: You can target 1 "Zoodiac" Xyz Monster you control; attach this card from your GY to that Xyz Monster as Xyz Material.',
    'Destroy as many Normal Monsters on the field as possible, and if you do, Special Summon Level 4 or lower Dinosaur-Type monsters from your Deck, up to the number destroyed, but destroy them during the End Phase. You can banish this card from your GY, then target 1 Dinosaur-Type monster you control and 1 card your opponent controls; destroy them.',
    'Target 1 "Raidraptor" monster you control; Special Summon 1 monster with the same name as that monster on the field from your hand or Deck in Defense Position. You can only activate 1 "Raidraptor - Call" per turn. You cannot Special Summon monsters during the turn you activate this card, except "Raidraptor" monsters.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 128]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Training Details

Training Dataset

pair_similarity_new_1231

Dataset: pair_similarity_new_1231 at e141fec
Size: 8,959 training samples
Columns: effect_text, score, and effect_text2
Approximate statistics based on the first 1000 samples:
effect_text score effect_text2
type string float string
details
min: 9 tokens
mean: 73.57 tokens
max: 204 tokens

min: 0.0
mean: 0.41
max: 1.0

min: 4 tokens
mean: 73.18 tokens
max: 203 tokens

	effect_text	score	effect_text2
type	string	float	string
details	min: 9 tokens mean: 73.57 tokens max: 204 tokens	min: 0.0 mean: 0.41 max: 1.0	min: 4 tokens mean: 73.18 tokens max: 203 tokens

Samples:

effect_text	score	effect_text2
`When your opponent's monster attacks a face-up Level 4 or lower Toon Monster on your side of the field, you can make the attack a direct attack to your Life Points.`	`0.0`	`During either player's Main Phase: Special Summon this card as a Normal Monster (Reptile-Type/EARTH/Level 4/ATK 1600/DEF 1800). (This card is also still a Trap Card.)`
`When your opponent Special Summons a monster, you can discard 1 card to Special Summon this card from your hand. Your opponent cannot remove cards from play.`	`1.0`	`Activate this card by discarding 1 monster, then target 1 monster in your GY whose Level is lower than the discarded monster's original Level; Special Summon it and equip it with this card. The equipped monster has its effects negated. You can only activate 1 "Overdone Burial" per turn.`
`Mystical Elf" + "Curtain of the Dark Ones`	`0.0`	`This card is used to Ritual Summon "Legendary Flame Lord". You must also Tribute monsters whose total Levels equal 7 or more from the field or your hand.`

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Evaluation Dataset

pair_similarity_new_1231

Dataset: pair_similarity_new_1231 at e141fec
Size: 1,920 evaluation samples
Columns: effect_text, score, and effect_text2
Approximate statistics based on the first 1000 samples:
effect_text score effect_text2
type string float string
details
min: 6 tokens
mean: 72.29 tokens
max: 190 tokens

min: 0.0
mean: 0.43
max: 1.0

min: 5 tokens
mean: 73.92 tokens
max: 206 tokens

	effect_text	score	effect_text2
type	string	float	string
details	min: 6 tokens mean: 72.29 tokens max: 190 tokens	min: 0.0 mean: 0.43 max: 1.0	min: 5 tokens mean: 73.92 tokens max: 206 tokens

Samples:

effect_text	score	effect_text2
`2+ Level 4 monsters This Xyz Summoned card gains 500 ATK x the total Link Rating of Link Monsters linked to this card. You can detach 2 materials from this card, then target 1 4 Cyberse Link Monster in your GY; Special Summon it to your field so it points to this card, also you cannot Special Summon other monsters or attack directly for the rest of this turn.`	`1.0`	3 Level 4 monsters Once per turn, you can also Xyz Summon "Zoodiac Tigermortar" by using 1 "Zoodiac" monster you control with a different name as Xyz Material. (If you used an Xyz Monster, any Xyz Materials attached to it also become Xyz Materials on this card.) This card gains ATK and DEF equal to the ATK and DEF of all "Zoodiac" monsters attached to it as Materials. Once per turn: You can detach 1 Xyz Material from this card, then target 1 Xyz Monster you control and 1 "Zoodiac" monster in your GY; attach that "Zoodiac" monster to that Xyz Monster as Xyz Material.
`1 Tuner + 1 or more non-Tuner Pendulum Monsters Once per turn: You can target 1 Pendulum Monster on the field or 1 card in the Pendulum Zone; destroy it, and if you do, shuffle 1 card on the field into the Deck. Once per turn: You can Special Summon 1 "Dracoslayer" monster from your Deck in Defense Position, but it cannot be used as a Synchro Material for a Summon.`	`0.5`	You can Ritual Summon this card with a "Recipe" card. If this card is Special Summoned: You can target 1 Spell/Trap on the field; destroy it. When a card or effect is activated that targets this card on the field, or when this card is targeted for an attack (Quick Effect): You can Tribute this card and 1 Attack Position monster on either field, and if you do, Special Summon 1 Level 3 or 4 "Nouvelles" Ritual Monster from your hand or Deck. You can only use each effect of "Confiras de Nouvelles" once per turn.
`If you control an Illusion or Spellcaster monster: Add 1 "White Forest" monster from your Deck to your hand. If this card is sent to the GY to activate a monster effect: You can Set this card. You can only use each effect of "Tales of the White Forest" once per turn.`	`1.0`	`If you control no monsters, you can Special Summon this card (from your hand). You can only use each of the following effects of "Kashtira Fenrir" once per turn. During your Main Phase: You can add 1 "Kashtira" monster from your Deck to your hand. When this card declares an attack, or if your opponent activates a monster effect (except during the Damage Step): You can target 1 face-up card your opponent controls; banish it, face-down.`

Loss: CoSENTLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "pairwise_cos_sim"
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
learning_rate: 1e-05
num_train_epochs: 10
warmup_ratio: 0.1
fp16: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 1e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 10
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: True
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
dispatch_batches: None
split_batches: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.1786	100	4.54	4.3773
0.3571	200	4.4771	4.3503
0.5357	300	4.4594	4.3149
0.7143	400	4.373	4.2905
0.8929	500	4.3383	4.2789
1.0714	600	4.3392	4.2705
1.25	700	4.3188	4.2627
1.4286	800	4.3372	4.2556
1.6071	900	4.3194	4.2487
1.7857	1000	4.2892	4.2420
1.9643	1100	4.2675	4.2375
2.1429	1200	4.3025	4.2312
2.3214	1300	4.2919	4.2246
2.5	1400	4.2853	4.2183
2.6786	1500	4.2463	4.2140
2.8571	1600	4.2486	4.2090
3.0357	1700	4.2376	4.2051
3.2143	1800	4.2365	4.1996
3.3929	1900	4.2223	4.1946
3.5714	2000	4.222	4.1900
3.75	2100	4.2057	4.1852
3.9286	2200	4.1935	4.1842
4.1071	2300	4.2001	4.1771
4.2857	2400	4.2012	4.1725
4.4643	2500	4.1961	4.1677
4.6429	2600	4.1919	4.1626
4.8214	2700	4.1542	4.1619
5.0	2800	4.1713	4.1594
5.1786	2900	4.1628	4.1530
5.3571	3000	4.1611	4.1504
5.5357	3100	4.1659	4.1480
5.7143	3200	4.131	4.1451
5.8929	3300	4.1191	4.1432
6.0714	3400	4.1521	4.1400
6.25	3500	4.1056	4.1374
6.4286	3600	4.1645	4.1340
6.6071	3700	4.1439	4.1305
6.7857	3800	4.0812	4.1298
6.9643	3900	4.0665	4.1323
7.1429	4000	4.1448	4.1261
7.3214	4100	4.1252	4.1236
7.5	4200	4.08	4.1234
7.6786	4300	4.0815	4.1188
7.8571	4400	4.085	4.1192
8.0357	4500	4.1015	4.1191
8.2143	4600	4.0912	4.1176
8.3929	4700	4.0822	4.1160
8.5714	4800	4.0744	4.1155
8.75	4900	4.1033	4.1135
8.9286	5000	4.0588	4.1135
9.1071	5100	4.0848	4.1125
9.2857	5200	4.0666	4.1125
9.4643	5300	4.0766	4.1125
9.6429	5400	4.0796	4.1116
9.8214	5500	4.0875	4.1117
10.0	5600	4.0929	4.1118

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.3.1
Transformers: 4.47.1
PyTorch: 2.5.1+cu121
Accelerate: 1.2.1
Datasets: 3.2.0
Tokenizers: 0.21.0

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

CoSENTLoss

@online{kexuefm-8847,
    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
    author={Su Jianlin},
    year={2022},
    month={Jan},
    url={https://kexue.fm/archives/8847},
}