test / README.md

Upload 11 files

d7d31ca verified 9 days ago

28.8 kB

language:
  - en
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:1375067
  - loss:MultipleNegativesRankingLoss
base_model: unsloth/all-MiniLM-L6-v2
widget:
  - source_sentence: |-
      Modify the inner parameters of the Kepler propagator in order to place
              the spacecraft in the right Sphere of Influence
    sentences:
      - "func (c *Conn) SetDeadline(t time.Time) error {\n\treturn c.p.SetDeadline(t)\n}"
      - |-
        def _change_soi(self, body):
                

                if body == self.central:
                    self.bodies = [self.central]
                    self.step = self.central_step
                    self.active = self.central.name
                    self.frame = self.central.name
                else:
                    soi = self.SOI[body.name]
                    self.bodies = [body]
                    self.step = self.alt_step
                    self.active = body.name
                    self.frame = soi.frame
      - |-
        def main(args=None):
            """"""
            parser = _parser()

            # Python 2 will error 'too few arguments' if no subcommand is supplied.
            # No such error occurs in Python 3, which makes it feasible to check
            # whether a subcommand was provided (displaying a help message if not).
            # argparse internals vary significantly over the major versions, so it's
            # much easier to just override the args passed to it. In this case, print
            # the usage message if there are no args.
            if args is None and len(sys.argv) <= 1:
                sys.argv.append('--help')

            options = parser.parse_args(args)

            # pass options to subcommand
            options.func(options)

            return 0
  - source_sentence: |-
      Load image from path.

      @param path Path to image.
      @return Image
      @throws java.io.IOException

      @throws NullPointerException if {@code path} is null.
    sentences:
      - "public function admin_modal_bail( $item_id, $item_title, $field_args ) {\n\n\t\t$model_data = $this->build_dfv_field_item_data_recurse_item( $item_id, $item_title, $field_args );\n\t\t?>\n\t\t\t<script type=\"text/javascript\">\n\t\t\t\twindow.parent.jQuery( window.parent ).trigger(\n\t\t\t\t\t'dfv:modal:update',\n\t\t\t\t\t<?php echo wp_json_encode( $model_data, JSON_HEX_TAG ); ?>\n\t\t\t\t);\n\t\t\t</script>\n\t\t<?php\n\n\t\tdie();\n\n\t}"
      - "private Image loadImage(Resource path) throws IOException {\n\t\tURL url = path.getURL();\n\t\tif (url == null) {\n\t\t\tlogger.warn(\"Unable to locate splash screen in classpath at: \" + path);\n\t\t\treturn null;\n\t\t}\n\t\treturn Toolkit.getDefaultToolkit().createImage(url);\n\t}"
      - |-
        def generate_wakeword_pieces(self, volume):
                """"""
                while True:
                    target = 1 if random() > 0.5 else 0
                    it = self.pos_files_it if target else self.neg_files_it
                    sample_file = next(it)
                    yield self.layer_with(self.normalize_volume_to(load_audio(sample_file), volume), target)
                    yield self.layer_with(np.zeros(int(pr.sample_rate * (0.5 + 2.0 * random()))), 0)
  - source_sentence: // StartPlugins starts all plugins in the correct order.
    sentences:
      - "func (co *Coordinator) StartPlugins() {\n\t// Launch routers\n\tfor _, router := range co.routers {\n\t\tlogrus.Debug(\"Starting \", reflect.TypeOf(router))\n\t\tif err := router.Start(); err != nil {\n\t\t\tlogrus.WithError(err).Errorf(\"Failed to start router of type '%s'\", reflect.TypeOf(router))\n\t\t}\n\t}\n\n\t// Launch producers\n\tco.state = coordinatorStateStartProducers\n\tfor _, producer := range co.producers {\n\t\tproducer := producer\n\t\tgo tgo.WithRecoverShutdown(func() {\n\t\t\tlogrus.Debug(\"Starting \", reflect.TypeOf(producer))\n\t\t\tproducer.Produce(co.producerWorker)\n\t\t})\n\t}\n\n\t// Set final log target and purge the intermediate buffer\n\tif core.StreamRegistry.IsStreamRegistered(core.LogInternalStreamID) {\n\t\t// The _GOLLUM_ stream has listeners, so use LogConsumer to write to it\n\t\tif *flagLogColors == \"always\" {\n\t\t\tlogrus.SetFormatter(logger.NewConsoleFormatter())\n\t\t}\n\t\tlogrusHookBuffer.SetTargetHook(co.logConsumer)\n\t\tlogrusHookBuffer.Purge()\n\n\t} else {\n\t\tlogrusHookBuffer.SetTargetWriter(logger.FallbackLogDevice)\n\t\tlogrusHookBuffer.Purge()\n\t}\n\n\t// Launch consumers\n\tco.state = coordinatorStateStartConsumers\n\tfor _, consumer := range co.consumers {\n\t\tconsumer := consumer\n\t\tgo tgo.WithRecoverShutdown(func() {\n\t\t\tlogrus.Debug(\"Starting \", reflect.TypeOf(consumer))\n\t\t\tconsumer.Consume(co.consumerWorker)\n\t\t})\n\t}\n}"
      - |-
        def __add_symbols(self, cmd):
                

                if self.__config.define_symbols:
                    symbols = self.__config.define_symbols
                    cmd.append(''.join(
                        [' -D"%s"' % def_symbol for def_symbol in symbols]))

                if self.__config.undefine_symbols:
                    un_symbols = self.__config.undefine_symbols
                    cmd.append(''.join(
                        [' -U"%s"' % undef_symbol for undef_symbol in un_symbols]))

                return cmd
      - |-
        protected function addReview()
            {
                if (!$this->isError()) {
                    $id = $this->review->add($this->getSubmitted());
                    if (empty($id)) {
                        $this->errorAndExit($this->text('Unexpected result'));
                    }
                    $this->line($id);
                }
            }
  - source_sentence: >-
      Modifies the result of each promise from a scalar value to a object
      containing its fieldname
    sentences:
      - "public void assertUniqueBeans(Set<String> ignoredDuplicateBeanNames) {\n\t\tfor (BeanohBeanFactoryMethodInterceptor callback : callbacks) {\n\t\t\tMap<String, List<BeanDefinition>> beanDefinitionMap = callback\n\t\t\t\t\t.getBeanDefinitionMap();\n\t\t\tfor (String key : beanDefinitionMap.keySet()) {\n\t\t\t\tif (!ignoredDuplicateBeanNames.contains(key)) {\n\t\t\t\t\tList<BeanDefinition> definitions = beanDefinitionMap\n\t\t\t\t\t\t\t.get(key);\n\t\t\t\t\tList<String> resourceDescriptions = new ArrayList<String>();\n\t\t\t\t\tfor (BeanDefinition definition : definitions) {\n\t\t\t\t\t\tString resourceDescription = definition\n\t\t\t\t\t\t\t\t.getResourceDescription();\n\t\t\t\t\t\tif (resourceDescription == null) {\n\t\t\t\t\t\t\tresourceDescriptions.add(definition.getBeanClassName());\n\t\t\t\t\t\t}else if (!resourceDescription\n\t\t\t\t\t\t\t\t.endsWith(\"-BeanohContext.xml]\")) {\n\t\t\t\t\t\t\tif(!resourceDescriptions.contains(resourceDescription)){\n\t\t\t\t\t\t\t\tresourceDescriptions.add(resourceDescription);\n\t\t\t\t\t\t\t}\n\t\t\t\t\t\t}\n\t\t\t\t\t}\n\t\t\t\t\tif (resourceDescriptions.size() > 1) {\n\t\t\t\t\t\tthrow new DuplicateBeanDefinitionException(\"Bean '\"\n\t\t\t\t\t\t\t\t+ key + \"' was defined \"\n\t\t\t\t\t\t\t\t+ resourceDescriptions.size() + \" times.\\n\"\n\t\t\t\t\t\t\t\t+ \"Either remove duplicate bean definitions or ignore them with the 'ignoredDuplicateBeanNames' method.\\n\"\n\t\t\t\t\t\t\t\t+ \"Configuration locations:\"\n\t\t\t\t\t\t\t\t+ messageUtil.list(resourceDescriptions));\n\t\t\t\t\t}\n\t\t\t\t}\n\t\t\t}\n\t\t}\n\t}"
      - |-
        function wrap(fieldName, promise, args) {
          return promise(args).then((result) => ({
            [fieldName]: result,
          }));
        }
      - "func Convert_kops_LyftVPCNetworkingSpec_To_v1alpha1_LyftVPCNetworkingSpec(in *kops.LyftVPCNetworkingSpec, out *LyftVPCNetworkingSpec, s conversion.Scope) error {\n\treturn autoConvert_kops_LyftVPCNetworkingSpec_To_v1alpha1_LyftVPCNetworkingSpec(in, out, s)\n}"
  - source_sentence: >-
      <p>

      User-supplied properties in key-value form.

      </p>


      @param parameters

      User-supplied properties in key-value form.

      @return Returns a reference to this object so that method calls can be
      chained together.
    sentences:
      - |-
        public static function unserializeFromStringRepresentation($string)
            {
                if (!preg_match('~k:(?P<k>\d+)/m:(?P<m>\d+)\((?P<bitfield>[0-9a-zA-Z+/=]+)\)~', $string, $matches)) {
                    throw new InvalidArgumentException('Invalid string representation');
                }
                $bf = new self((int) $matches['m'], (int) $matches['k']);
                $bf->bitField = base64_decode($matches['bitfield']);
                return $bf;
            }
      - |-
        public static function flushEventListeners()
            {
                if (! isset(static::$dispatcher)) {
                    return;
                }

                $instance = new static;

                foreach ($instance->getObservableEvents() as $event) {
                    static::$dispatcher->forget("eloquent.{$event}: ".static::class);
                }

                foreach (array_values($instance->dispatchesEvents) as $event) {
                    static::$dispatcher->forget($event);
                }
            }
      - >-
        public StorageDescriptor withParameters(java.util.Map<String, String>
        parameters) {
                setParameters(parameters);
                return this;
            }
datasets:
  - sentence-transformers/codesearchnet
pipeline_tag: sentence-similarity
library_name: sentence-transformers

SentenceTransformer based on unsloth/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from unsloth/all-MiniLM-L6-v2 on the codesearchnet dataset. It maps sentences & paragraphs to a 384-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: unsloth/all-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- codesearchnet
Language: en

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': 256, 'do_lower_case': False, 'architecture': 'PeftModelForFeatureExtraction'})
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (2): Normalize()
)

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("sentence_transformers_model_id")
# Run inference
sentences = [
    '<p>\nUser-supplied properties in key-value form.\n</p>\n\n@param parameters\nUser-supplied properties in key-value form.\n@return Returns a reference to this object so that method calls can be chained together.',
    'public StorageDescriptor withParameters(java.util.Map<String, String> parameters) {\n        setParameters(parameters);\n        return this;\n    }',
    "public static function unserializeFromStringRepresentation($string)\n    {\n        if (!preg_match('~k:(?P<k>\\d+)/m:(?P<m>\\d+)\\((?P<bitfield>[0-9a-zA-Z+/=]+)\\)~', $string, $matches)) {\n            throw new InvalidArgumentException('Invalid string representation');\n        }\n        $bf = new self((int) $matches['m'], (int) $matches['k']);\n        $bf->bitField = base64_decode($matches['bitfield']);\n        return $bf;\n    }",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.6597, -0.0469],
#         [ 0.6597,  1.0000,  0.0107],
#         [-0.0469,  0.0107,  1.0000]], dtype=torch.float16)

Training Details

Training Dataset

codesearchnet

Dataset: codesearchnet at 079a958
Size: 1,375,067 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 4 tokens
mean: 29.95 tokens
max: 127 tokens

min: 28 tokens
mean: 131.03 tokens
max: 256 tokens

	anchor	positive
type	string	string
details	min: 4 tokens mean: 29.95 tokens max: 127 tokens	min: 28 tokens mean: 131.03 tokens max: 256 tokens

Samples:

anchor	positive
`Computes the new parent id for the node being moved. @return int`	`protected function parentId() { switch ( $this->position ) { case 'root': return null; case 'child': return $this->target->getKey(); default: return $this->target->getParentId(); } }`
`// SetWinSize overwrites the playlist's window size.`	`func (p *MediaPlaylist) SetWinSize(winsize uint) error { if winsize > p.capacity { return errors.New("capacity must be greater than winsize or equal") } p.winsize = winsize return nil }`
`Show the sidebar and squish the container to make room for the sidebar. If hideOthers is true, hide other open sidebars.`	`function() { var options = this.options; if (options.hideOthers) { this.secondary.each(function() { var sidebar = $(this); if (sidebar.hasClass('is-expanded')) { sidebar.toolkit('offCanvas', 'hide'); } }); } this.fireEvent('showing'); this.container.addClass('move-' + this.opposite); this.element .reveal() .addClass('is-expanded') .aria('expanded', true); if (options.stopScroll) { $('body').addClass('no-scroll'); } this.fireEvent('shown'); }`

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 64
gradient_accumulation_steps: 4
learning_rate: 0.0002
num_train_epochs: 2
warmup_ratio: 0.03
fp16: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: no
prediction_loss_only: True
per_device_train_batch_size: 64
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 4
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 0.0002
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 2
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.03
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: 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}
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: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss
0.0186	50	0.5333
0.0372	100	0.3948
0.0559	150	0.311
0.0745	200	0.2721
0.0931	250	0.2809
0.1117	300	0.2533
0.1303	350	0.2472
0.1489	400	0.2378
0.1676	450	0.2383
0.1862	500	0.2239
0.2048	550	0.2236
0.2234	600	0.2191
0.2420	650	0.2248
0.2606	700	0.2176
0.2793	750	0.2171
0.2979	800	0.2114
0.3165	850	0.222
0.3351	900	0.2066
0.3537	950	0.2059
0.3723	1000	0.2053
0.3910	1050	0.2011
0.4096	1100	0.2024
0.4282	1150	0.2006
0.4468	1200	0.1976
0.4654	1250	0.1968
0.4840	1300	0.195
0.5027	1350	0.1921
0.5213	1400	0.1967
0.5399	1450	0.1895
0.5585	1500	0.1864
0.5771	1550	0.189
0.5957	1600	0.1857
0.6144	1650	0.1889
0.6330	1700	0.1796
0.6516	1750	0.1718
0.6702	1800	0.1866
0.6888	1850	0.1874
0.7074	1900	0.178
0.7261	1950	0.1763
0.7447	2000	0.1734
0.7633	2050	0.1823
0.7819	2100	0.1796
0.8005	2150	0.1737
0.8191	2200	0.1796
0.8378	2250	0.1794
0.8564	2300	0.1703
0.8750	2350	0.1746
0.8936	2400	0.1864
0.9122	2450	0.173
0.9308	2500	0.1729
0.9495	2550	0.1742
0.9681	2600	0.1776
0.9867	2650	0.182
1.0052	2700	0.1661
1.0238	2750	0.1627
1.0424	2800	0.158
1.0611	2850	0.1585
1.0797	2900	0.1555
1.0983	2950	0.1566
1.1169	3000	0.1511
1.1355	3050	0.1557
1.1541	3100	0.1589
1.1728	3150	0.1545
1.1914	3200	0.1567
1.2100	3250	0.1561
1.2286	3300	0.1515
1.2472	3350	0.153
1.2658	3400	0.1557
1.2845	3450	0.1506
1.3031	3500	0.1572
1.3217	3550	0.1543
1.3403	3600	0.1619
1.3589	3650	0.1586
1.3775	3700	0.16
1.3962	3750	0.1594
1.4148	3800	0.1528
1.4334	3850	0.1516
1.4520	3900	0.1529
1.4706	3950	0.149
1.4892	4000	0.1572
1.5079	4050	0.1505
1.5265	4100	0.1552
1.5451	4150	0.1488
1.5637	4200	0.161
1.5823	4250	0.151
1.6009	4300	0.1442
1.6196	4350	0.1511
1.6382	4400	0.1475
1.6568	4450	0.1509
1.6754	4500	0.1512
1.6940	4550	0.1484
1.7127	4600	0.1491
1.7313	4650	0.143
1.7499	4700	0.1479
1.7685	4750	0.1459
1.7871	4800	0.1434
1.8057	4850	0.1475
1.8244	4900	0.1485
1.8430	4950	0.147
1.8616	5000	0.157
1.8802	5050	0.1447
1.8988	5100	0.1425
1.9174	5150	0.1491
1.9361	5200	0.1433
1.9547	5250	0.1382
1.9733	5300	0.1391
1.9919	5350	0.1492

Framework Versions

Python: 3.12.12
Sentence Transformers: 5.1.1
Transformers: 4.57.1
PyTorch: 2.10.0+cu128
Accelerate: 1.11.0
Datasets: 4.3.0
Tokenizers: 0.22.1

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",
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}