This is a sentence-transformers model finetuned from facebook/esm2_t6_8M_UR50D. It maps sentences & paragraphs to a 320-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: facebook/esm2_t6_8M_UR50D
Maximum Sequence Length: 1026 tokens
Output Dimensionality: 320 dimensions
Similarity Function: Cosine Similarity

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': 1026, 'do_lower_case': False, 'architecture': 'EsmModel'})
  (1): Pooling({'word_embedding_dimension': 320, '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("HassanCS/peptide_HLA_TCRa_TCRb__esm2_t6_8M_UR50D_up_to_epoch_8")
# Run inference
sentences = [
    'M E V T P S G T W L',
    'G Q A R V A Y Q V',
    'L L L A R A A S L S L',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 320]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.4961, 0.5182],
#         [0.4961, 1.0000, 0.8793],
#         [0.5182, 0.8793, 1.0000]])

Evaluation

Metrics

Semantic Similarity

Dataset: all-dev
Evaluated with EmbeddingSimilarityEvaluator

Metric	Value
pearson_cosine	0.9649
spearman_cosine	0.9629

Training Details

Training Dataset

Unnamed Dataset

Size: 101,745 training samples
Columns: sentence1, sentence2, and score
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 score
type string string float
details
min: 9 tokens
mean: 11.56 tokens
max: 26 tokens
min: 10 tokens
mean: 11.47 tokens
max: 26 tokens
min: 0.03
mean: 0.3
max: 1.0
Samples:
sentence1 sentence2 score
V Y G I R L E H F A L G W V F V P V 0.4487163971048506
G R I A F F L K Y I P S I N V H H Y 0.2950890611282742
I L A K F A H W L Q T N P V T L Q Y 0.24611965364324553
Loss: CoSENTLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "pairwise_cos_sim" 4}

Evaluation Dataset

Unnamed Dataset

Size: 11,305 evaluation samples
Columns: sentence1, sentence2, and score
Approximate statistics based on the first 1000 samples:
sentence1 sentence2 score
type string string float
details
min: 10 tokens
mean: 11.48 tokens
max: 26 tokens
min: 9 tokens
mean: 11.47 tokens
max: 26 tokens
min: 0.02
mean: 0.31
max: 1.0
Samples:
sentence1 sentence2 score
K S K R T P M G F G A D G V G K S A L 0.35084774335438396
T P R V T G G G A M E I F D R Y G E E V 0.21484461945678654
S R I M L L A P K K I F G S L A F L 0.14818627314226981
Loss: CoSENTLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "pairwise_cos_sim" 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
learning_rate: 0.001
weight_decay: 0.0001
num_train_epochs: 8
fp16: True
load_best_model_at_end: True

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 128
per_device_eval_batch_size: 128
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: 0.001
weight_decay: 0.0001
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 8
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
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
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: True
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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
hub_revision: None
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
include_tokens_per_second: False
include_num_input_tokens_seen: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	all-dev_spearman_cosine
0.1258	100	8.9433	-	-
0.2516	200	8.7471	-	-
0.3774	300	8.6225	-	-
0.5031	400	8.5602	-	-
0.6289	500	8.5013	-	-
0.7547	600	8.4547	-	-
0.8805	700	8.4103	-	-
1.0	795	-	8.3649	0.8817
1.0063	800	8.3797	-	-
1.1321	900	8.3271	-	-
1.2579	1000	8.3076	-	-
1.3836	1100	8.2519	-	-
1.5094	1200	8.2203	-	-
1.6352	1300	8.2108	-	-
1.7610	1400	8.1601	-	-
1.8868	1500	8.1745	-	-
2.0	1590	-	8.1604	0.9278
2.0126	1600	8.1477	-	-
2.1384	1700	8.192	-	-
2.2642	1800	8.1813	-	-
2.3899	1900	8.1643	-	-
2.5157	2000	8.1401	-	-
2.6415	2100	8.1043	-	-
2.7673	2200	8.1092	-	-
2.8931	2300	8.0935	-	-
3.0	2385	-	8.0852	0.9385
3.0189	2400	8.0631	-	-
3.1447	2500	8.031	-	-
3.2704	2600	8.0052	-	-
3.3962	2700	7.9924	-	-
3.5220	2800	7.9701	-	-
3.6478	2900	7.958	-	-
3.7736	3000	7.9537	-	-
3.8994	3100	7.9321	-	-
4.0	3180	-	7.9955	0.9511
4.0252	3200	7.9196	-	-
4.1509	3300	7.9891	-	-
4.2767	3400	7.9788	-	-
4.4025	3500	7.9793	-	-
4.5283	3600	7.9699	-	-
4.6541	3700	7.9371	-	-
4.7799	3800	7.9548	-	-
4.9057	3900	7.9349	-	-
5.0	3975	-	7.9789	0.9520
5.0314	4000	7.9036	-	-
5.1572	4100	7.874	-	-
5.2830	4200	7.853	-	-
5.4088	4300	7.8523	-	-
5.5346	4400	7.8475	-	-
5.6604	4500	7.8436	-	-
5.7862	4600	7.8402	-	-
5.9119	4700	7.8097	-	-
6.0	4770	-	7.9175	0.9592
6.0377	4800	7.8271	-	-
6.1635	4900	7.8704	-	-
6.2893	5000	7.8793	-	-
6.4151	5100	7.8644	-	-
6.5409	5200	7.8611	-	-
6.6667	5300	7.8571	-	-
6.7925	5400	7.8439	-	-
6.9182	5500	7.845	-	-
7.0	5565	-	7.9248	0.9585
7.0440	5600	7.795	-	-
7.1698	5700	7.7812	-	-
7.2956	5800	7.7793	-	-
7.4214	5900	7.7608	-	-
7.5472	6000	7.7587	-	-
7.6730	6100	7.7586	-	-
7.7987	6200	7.7612	-	-
7.9245	6300	7.7674	-	-
8.0	6360	-	7.8833	0.9629

The bold row denotes the saved checkpoint.

Framework Versions

Python: 3.10.19
Sentence Transformers: 5.1.2
Transformers: 4.57.1
PyTorch: 2.9.1+cu128
Accelerate: 1.11.0
Datasets: 4.4.1
Tokenizers: 0.22.1

Citation

BibTeX

Sentence Transformers

bibtex

@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

bibtex

@article{10531646,
    author={Huang, Xiang and Peng, Hao and Zou, Dongcheng and Liu, Zhiwei and Li, Jianxin and Liu, Kay and Wu, Jia and Su, Jianlin and Yu, Philip S.},
    journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
    title={CoSENT: Consistent Sentence Embedding via Similarity Ranking},
    year={2024},
    doi={10.1109/TASLP.2024.3402087}
}

	sentence1	sentence2	score
type	string	string	float
details	min: 9 tokens mean: 11.56 tokens max: 26 tokens	min: 10 tokens mean: 11.47 tokens max: 26 tokens	min: 0.03 mean: 0.3 max: 1.0

sentence1	sentence2	score
`V Y G I R L E H F`	`A L G W V F V P V`	`0.4487163971048506`
`G R I A F F L K Y`	`I P S I N V H H Y`	`0.2950890611282742`
`I L A K F A H W L`	`Q T N P V T L Q Y`	`0.24611965364324553`

sentence1	sentence2	score
`K S K R T P M G F`	`G A D G V G K S A L`	`0.35084774335438396`
`T P R V T G G G A M`	`E I F D R Y G E E V`	`0.21484461945678654`
`S R I M L L A P K`	`K I F G S L A F L`	`0.14818627314226981`