Update README.md

README.md

@@ -1,621 +1,123 @@
 ---
 tags:
-- sentence-transformers
-- sentence-similarity
-- feature-extraction
-- generated_from_trainer
-- dataset_size:2022217
-- loss:MultipleNegativesRankingLoss
+- code
+- python
+- php
+- java
+- javascript
+- go
+- ruby
+- rust
 base_model: Shuu12121/CodeModernBERT-Crow
-widget:
-- source_sentence: 'Clone value to a new instance
-
-
-    @private
-
-    @param {*} val
-
-    @returns {*}'
-  sentences:
-  - "function _copy(val) {\n    const type = $type(val);\n\n    if (type == 'object')\
-    \ {\n        val = _extend({}, val, true);\n    } else if (type == 'array') {\n\
-    \        val = val.slice(0);\n    }\n\n    return val;\n}"
-  - "function (data) {\n    data = data || {};\n\n    this.category = data.hasOwnProperty('category')\
-    \ ? data.category : 'No category';\n    this.id = data.hasOwnProperty('id') ?\
-    \ data.id : '';\n    this.group = data.hasOwnProperty('group') ? data.group :\
-    \ '';\n    this.lines = data.hasOwnProperty('lines') ? data.lines : 0;\n    this.name\
-    \ = data.hasOwnProperty('name') ? data.name : '';\n    this.options = data.hasOwnProperty('options')\
-    \ ? data.options : {};\n    this.origin = data.hasOwnProperty('origin') ? data.origin\
-    \ : '';\n    this.resources = data.hasOwnProperty('resources') ? data.resources\
-    \ : Object.assign({}, getDefaultOptions().resources);\n    this.usage = data.hasOwnProperty('usage')\
-    \ ? data.usage : [];\n    this.viewId = data.hasOwnProperty('viewId') ? data.viewId\
-    \ : '';\n}"
-  - "public function relativeMove($model, $position)\n    {\n        $conditionAttributes\
-    \ = (array)$this->conditionAttributes;\n        $owner = $this->owner;\n\n   \
-    \     if (!empty($conditionAttributes)) {\n            $sameCondition = true;\n\
-    \            foreach ($conditionAttributes as $attr) {\n                if ($owner->getAttribute($attr)\
-    \ != $model->getAttribute($attr)) {\n                    $sameCondition = false;\n\
-    \                    break;\n                }\n            }\n            if\
-    \ (!$sameCondition) {\n                // move in other condition category\n \
-    \               $this->moveToTop();\n                // update condition attribute\n\
-    \                $condition = [];\n                foreach ($conditionAttributes\
-    \ as $attr) {\n                    $condition[$attr] = $model->getAttribute($attr);\n\
-    \                }\n                $condition[$this->sortAttribute] = $owner->find()->andWhere($this->getCondition())->count()\
-    \ - 1;\n                $owner->updateAttributes($condition);\n            }\n\
-    \        }\n        // calculate pos change\n        $currentPos = $owner->getAttribute($this->sortAttribute);\n\
-    \        $destinationPos = $model->getAttribute($this->sortAttribute);\n     \
-    \   if ($position == 'after') {\n            $newPos = $destinationPos > $currentPos\
-    \ ? $destinationPos - 1 : $destinationPos;\n        } else {\n            $newPos\
-    \ = $destinationPos > $currentPos ? $destinationPos : $destinationPos + 1;\n \
-    \       }\n        $this->moveToPosition($newPos);\n    }"
-- source_sentence: '/*
-
-    Realize an asynchronous squeue command on slurm according a parameter (or not).
-
-    Data are formated into a literal.
-
-    @paramSqueue {string} optional. For example : '' -o "%j %i" '' // not implemented
-    yet'
-  sentences:
-  - "function(paramSqueue) {\n    if (!paramSqueue) paramSqueue = '';\n    paramSqueue\
-    \ = ''; // to remove when it will be take into account in the implementation\n\
-    \    var emitter = new events.EventEmitter();\n    var squeueRes_dict = {\n  \
-    \      'id': [],\n        'partition': [],\n        'nameUUID': [],\n        'status':\
-    \ []\n    }\n\n    // squeue command\n    var exec_cmd = require('child_process').exec;\n\
-    \    exec_cmd(queueBinary + '  -o \\\"\\%i \\%P \\%j \\%t\\\" ' + paramSqueue,\
-    \ function(err, stdout, stderr) {\n        if (err) {\n            emitter.emit('listError',\
-    \ err);\n            return;\n        }\n        var squeueRes_str = ('' + stdout).replace(/\\\
-    \"/g, ''); // squeue results\n        squeueRes_str.split('\\n')\n           \
-    \ .filter(function(jobArray, i) {\n                return jobArray.length > 0\
-    \ && i > 0;\n            })\n            .map(function(jobLine, i) { // for each\
-    \ job\n                return test = jobLine.split(' ').filter(function(val) {\n\
-    \                    return val != ''; // keep values that are not empty\n   \
-    \             });\n            })\n            .map(function(jobArray, i) { //\
-    \ save each field in the corresponding array of dict\n                squeueRes_dict.id.push(jobArray[0]);\
-    \ // job ID gived by slurm\n                squeueRes_dict.partition.push(jobArray[1]);\
-    \ // gpu, cpu, etc.\n                squeueRes_dict.nameUUID.push(jobArray[2]);\
-    \ // unique job ID gived by Nslurm (uuid)\n                squeueRes_dict.status.push(jobArray[3]);\
-    \ // P, R, CF, CG, etc.\n            });\n        emitter.emit('data', squeueRes_dict);\n\
-    \    });\n    return emitter;\n}"
-  - "function(node, state, leaving) {\n        for (let i = 0; i < state.currentSegments.length;\
-    \ ++i) {\n            const segInternal = state.currentSegments[i].internal;\n\
-    \n            if (leaving) {\n                segInternal.exitNodes.push(node);\n\
-    \            } else {\n                segInternal.nodes.push(node);\n       \
-    \     }\n        }\n\n        debug([\n            `${state.currentSegments.map(getId).join(\"\
-    ,\")})`,\n            `${node.type}${leaving ? \":exit\" : \"\"}`\n        ].join(\"\
-    \ \"));\n    }"
-  - "private void checkForGenerator(final Class<?> clazz, Field field, GeneratedValue\
-    \ generatedValue, String schemaName)\n\n    {\n\n        TableGenerator tableGenerator\
-    \ = field.getAnnotation(TableGenerator.class);\n        SequenceGenerator sequenceGenerator\
-    \ = field.getAnnotation(SequenceGenerator.class);\n        if (tableGenerator\
-    \ == null || !tableGenerator.name().equals(generatedValue.generator()))\n    \
-    \    {\n            tableGenerator = clazz.getAnnotation(TableGenerator.class);\n\
-    \        }\n        if (sequenceGenerator == null || !sequenceGenerator.name().equals(generatedValue.generator()))\n\
-    \        {\n            sequenceGenerator = clazz.getAnnotation(SequenceGenerator.class);\n\
-    \        }\n\n        if ((tableGenerator == null && sequenceGenerator == null)\n\
-    \                || (tableGenerator != null && !tableGenerator.name().equals(generatedValue.generator()))\n\
-    \                || (sequenceGenerator != null && !sequenceGenerator.name().equals(generatedValue.generator())))\n\
-    \        {\n\n            throw new RuleValidationException(\"Unknown Id.generator:\
-    \ \" + generatedValue.generator());\n\n        }\n        else if ((tableGenerator\
-    \ != null && !tableGenerator.schema().isEmpty() && !tableGenerator.schema().equals(\n\
-    \                schemaName))\n                || (sequenceGenerator != null &&\
-    \ !sequenceGenerator.schema().isEmpty() && !sequenceGenerator.schema()\n     \
-    \                   .equals(schemaName)))\n        {\n\n            throw new\
-    \ RuleValidationException(\"Generator \" + generatedValue.generator() + \" in\
-    \ entity : \"\n                    + clazz.getName() + \" has different schema\
-    \ name ,it should be same as entity have\");\n\n        }\n\n    }"
-- source_sentence: '@param      $param1
-
-    @param null $param2
-
-    @param null $param3
-
-
-    @return array|int|mixed
-
-    @throws DbException'
-  sentences:
-  - "func (s *GetReservationUtilizationOutput) SetUtilizationsByTime(v []*UtilizationByTime)\
-    \ *GetReservationUtilizationOutput {\n\ts.UtilizationsByTime = v\n\treturn s\n\
-    }"
-  - "public static function ask($param1, $param2 = null, $param3 = null) {\n     \
-    \   self::init();\n        if(is_array($param1)) {\n            return self::smartSelect($param1,\
-    \ $param2, $param3);\n        } else {\n            switch(substr($param1, 0,\
-    \ 1)) {\n                case '>':\n                case '/':\n              \
-    \      return self::smartQuery($param1, $param2);\n                    break;\n\
-    \                case '?':\n                    return self::smartSelect(substr($param1,\
-    \ 1), $param2, $param3);\n                    break;\n                default:\n\
-    \                    if(is_array($param3)) {\n                        return self::smartUpdate($param1,\
-    \ $param2, $param3);\n                    } else {\n                        return\
-    \ self::smartInsert($param1, $param2);\n                    }\n            }\n\
-    \        }\n    }"
-  - "void printStates() {\n           int     c;    // input \"character\"\n     \
-    \      int     n;    // state number\n\n           System.out.print(\"state |\
-    \           i n p u t     s y m b o l s \\n\");\n           System.out.print(\"\
-    \      | Acc  LA    Tag\");\n           for (c=0; c<fRB.fSetBuilder.getNumCharCategories();\
-    \ c++) {\n               RBBINode.printInt(c, 3);\n           }\n           System.out.print(\"\
-    \\n\");\n           System.out.print(\"      |---------------\");\n          \
-    \ for (c=0; c<fRB.fSetBuilder.getNumCharCategories(); c++) {\n               System.out.print(\"\
-    ---\");\n           }\n           System.out.print(\"\\n\");\n\n           for\
-    \ (n=0; n<fDStates.size(); n++) {\n               RBBIStateDescriptor sd = fDStates.get(n);\n\
-    \               RBBINode.printInt(n, 5);\n               System.out.print(\" |\
-    \ \");\n\n               RBBINode.printInt(sd.fAccepting, 3);\n              \
-    \ RBBINode.printInt(sd.fLookAhead, 4);\n               RBBINode.printInt(sd.fTagsIdx,\
-    \ 6);\n               System.out.print(\" \");\n               for (c=0; c<fRB.fSetBuilder.getNumCharCategories();\
-    \ c++) {\n                   RBBINode.printInt(sd.fDtran[c], 3);\n           \
-    \    }\n               System.out.print(\"\\n\");\n           }\n           System.out.print(\"\
-    \\n\\n\");\n       }"
-- source_sentence: 'Performs a forward, allowing page-relative paths and setting all
-    values
-
-    compatible with &lt;ao:forward&gt; tag.
-
-
-    @param  args  The arguments for the page, make unmodifiable and accessible as
-    request-scope var "arg"
-
-
-    @see #forward(java.lang.String, javax.servlet.RequestDispatcher, javax.servlet.http.HttpServletRequest,
-    javax.servlet.http.HttpServletResponse, java.util.Map)'
-  sentences:
-  - "public static void forward(\n\t\tServletContext servletContext,\n\t\tString page,\n\
-    \t\tHttpServletRequest request,\n\t\tHttpServletResponse response,\n\t\tMap<String,?>\
-    \ args\n\t) throws ServletException, IOException {\n\t\t// Resolve the dispatcher\n\
-    \t\tString contextRelativePath = ServletUtil.getAbsolutePath(getCurrentPagePath(request),\
-    \ page);\n\t\tRequestDispatcher dispatcher = servletContext.getRequestDispatcher(contextRelativePath);\n\
-    \t\tif(dispatcher==null) throw new LocalizedServletException(accessor, \"Dispatcher.dispatcherNotFound\"\
-    , contextRelativePath);\n\t\tforward(contextRelativePath, dispatcher, request,\
-    \ response, args);\n\t}"
-  - "public function getConnection($name = ''): QueryBuilderInterface\n\t{\n\t\t//\
-    \ If the parameter is a string, use it as an array index\n\t\tif (is_scalar($name)\
-    \ && isset($this->connections[$name]))\n\t\t{\n\t\t\treturn $this->connections[$name];\n\
-    \t\t}\n\t\telse if (empty($name) && ! empty($this->connections)) // Otherwise,\
-    \ return the last one\n\t\t{\n\t\t\treturn end($this->connections);\n\t\t}\n\n\
-    \t\t// You should actually connect before trying to get a connection...\n\t\t\
-    throw new InvalidArgumentException('The specified connection does not exist');\n\
-    \t}"
-  - "func (c *Context) Untrack(class, id string) error {\n\tfullID := payload.BuildID(class,\
-    \ id)\n\tlogger.Tracef(\"Calling untrack on payload context %q\", fullID)\n\n\t\
-    res, err := c.api.Untrack(fullID)\n\tif err != nil {\n\t\treturn errors.Trace(err)\n\
-    \t}\n\t// TODO(ericsnow) We should not ignore a 0-len result.\n\tif len(res) >\
-    \ 0 && res[0].Error != nil {\n\t\treturn errors.Trace(res[0].Error)\n\t}\n\tdelete(c.payloads,\
-    \ id)\n\n\treturn nil\n}"
-- source_sentence: /* PRIVATE
-  sentences:
-  - "void activate(ProtocolVersion helloVersion) throws IOException {\n        if\
-    \ (activeProtocols == null) {\n            activeProtocols = getActiveProtocols();\n\
-    \        }\n\n        if (activeProtocols.collection().isEmpty() ||\n        \
-    \        activeProtocols.max.v == ProtocolVersion.NONE.v) {\n            throw\
-    \ new SSLHandshakeException(\n                    \"No appropriate protocol (protocol\
-    \ is disabled or \" +\n                    \"cipher suites are inappropriate)\"\
-    );\n        }\n\n        if (activeCipherSuites == null) {\n            activeCipherSuites\
-    \ = getActiveCipherSuites();\n        }\n\n        if (activeCipherSuites.collection().isEmpty())\
-    \ {\n            throw new SSLHandshakeException(\"No appropriate cipher suite\"\
-    );\n        }\n\n        // temporary protocol version until the actual protocol\
-    \ version\n        // is negotiated in the Hello exchange. This affects the record\n\
-    \        // version we sent with the ClientHello.\n        if (!isInitialHandshake)\
-    \ {\n            protocolVersion = activeProtocolVersion;\n        } else {\n\
-    \            protocolVersion = activeProtocols.max;\n        }\n\n        if (helloVersion\
-    \ == null || helloVersion.v == ProtocolVersion.NONE.v) {\n            helloVersion\
-    \ = activeProtocols.helloVersion;\n        }\n\n        // We accumulate digests\
-    \ of the handshake messages so that\n        // we can read/write CertificateVerify\
-    \ and Finished messages,\n        // getting assurance against some particular\
-    \ active attacks.\n        Set<String> localSupportedHashAlgorithms =\n      \
-    \      SignatureAndHashAlgorithm.getHashAlgorithmNames(\n                getLocalSupportedSignAlgs());\n\
-    \        handshakeHash = new HandshakeHash(!isClient, needCertVerify,\n      \
-    \      localSupportedHashAlgorithms);\n\n        // Generate handshake input/output\
-    \ stream.\n        input = new HandshakeInStream(handshakeHash);\n        if (conn\
-    \ != null) {\n            output = new HandshakeOutStream(protocolVersion, helloVersion,\n\
-    \                                        handshakeHash, conn);\n            conn.getAppInputStream().r.setHandshakeHash(handshakeHash);\n\
-    \            conn.getAppInputStream().r.setHelloVersion(helloVersion);\n     \
-    \       conn.getAppOutputStream().r.setHelloVersion(helloVersion);\n        }\
-    \ else {\n            output = new HandshakeOutStream(protocolVersion, helloVersion,\n\
-    \                                        handshakeHash, engine);\n           \
-    \ engine.inputRecord.setHandshakeHash(handshakeHash);\n            engine.inputRecord.setHelloVersion(helloVersion);\n\
-    \            engine.outputRecord.setHelloVersion(helloVersion);\n        }\n\n\
-    \        // move state to activated\n        state = -1;\n    }"
-  - "function _error(coreIndex, cores) {\n  var errMsg =\n    '[cpu-stats] Error:\
-    \ Core \"' + coreIndex + '\" not found, use one of ' +\n    '[0, ' + (cores -\
-    \ 1) + '], ' +\n    'since your system has a total of ' + cores + ' cores.';\n\
-    \  console.log(errMsg);\n}"
-  - "function _drawLine(v0, v1, color) {\n        var p = new Primitive();\n\n   \
-    \     p.vertices = [v0, v1];\n        p.color = toColor(color);\n\n        renderer.addPrimitive(p);\n\
-    \    }"
 pipeline_tag: sentence-similarity
 library_name: sentence-transformers
+license: apache-2.0
+datasets:
+- Shuu12121/python-codesearch-filtered
+- Shuu12121/ruby-codesearch-filtered
+- Shuu12121/java-codesearch-filtered
+- Shuu12121/go-codesearch-filtered
+- Shuu12121/rust-codesearch-filtered
+- Shuu12121/javascript-codesearch-filtered
+- Shuu12121/php-codesearch-filtered
+- code-search-net/code_search_net
+language:
+- en
 ---
+# Shuu12121/CodeSearch-ModernBERT-Crow-Plus-1.0 🐦‍⬛
 
-# SentenceTransformer based on Shuu12121/CodeModernBERT-Crow
-
-This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [Shuu12121/CodeModernBERT-Crow](https://huggingface.co/Shuu12121/CodeModernBERT-Crow). It maps sentences & paragraphs to a 768-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:** [Shuu12121/CodeModernBERT-Crow](https://huggingface.co/Shuu12121/CodeModernBERT-Crow) <!-- at revision fa8af97f22c9bf631506435c994e724f82747e67 -->
-- **Maximum Sequence Length:** 1024 tokens
-- **Output Dimensionality:** 768 dimensions
-- **Similarity Function:** Cosine Similarity
-<!-- - **Training Dataset:** Unknown -->
-<!-- - **Language:** Unknown -->
-<!-- - **License:** Unknown -->
-
-### Model Sources
-
-- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
-- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
-- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
-
-### Full Model Architecture
-
-```
-SentenceTransformer(
-  (0): Transformer({'max_seq_length': 1024, 'do_lower_case': False}) with Transformer model: ModernBertModel 
-  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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:
-
-```bash
-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("sentence_transformers_model_id")
-# Run inference
-sentences = [
-    '/* PRIVATE',
-    'function _error(coreIndex, cores) {\n  var errMsg =\n    \'[cpu-stats] Error: Core "\' + coreIndex + \'" not found, use one of \' +\n    \'[0, \' + (cores - 1) + \'], \' +\n    \'since your system has a total of \' + cores + \' cores.\';\n  console.log(errMsg);\n}',
-    'function _drawLine(v0, v1, color) {\n        var p = new Primitive();\n\n        p.vertices = [v0, v1];\n        p.color = toColor(color);\n\n        renderer.addPrimitive(p);\n    }',
-]
-embeddings = model.encode(sentences)
-print(embeddings.shape)
-# [3, 768]
-
-# Get the similarity scores for the embeddings
-similarities = model.similarity(embeddings, embeddings)
-print(similarities.shape)
-# [3, 3]
-```
-
-<!--
-### Direct Usage (Transformers)
-
-<details><summary>Click to see the direct usage in Transformers</summary>
+このモデルは、`Shuu12121/CodeModernBERT-Crow` をベースにした Sentence Transformer モデルであり、**PHPとGoのクリーン版データセット**を新たに加えることで、さらに高精度な多言語コード検索を実現しています。  
+本バージョン（1.0）では、前バージョン (`Crow-Plus`) に対して若干の性能向上が確認されています。
 
-</details>
--->
+> This is an enhanced version of `CodeSearch-ModernBERT-Crow-Plus`, based on `CodeModernBERT-Crow`, incorporating clean PHP and Go datasets for improved multilingual code search performance.
 
-<!--
-### Downstream Usage (Sentence Transformers)
+開発者 (Developer): [Shuu12121](https://huggingface.co/Shuu12121)  
+ベースモデル (Base Model): [Shuu12121/CodeModernBERT-Crow](https://huggingface.co/Shuu12121/CodeModernBERT-Crow)  
+ライセンス (License): Apache-2.0
 
-You can finetune this model on your own dataset.
-
-<details><summary>Click to expand</summary>
-
-</details>
--->
-
-<!--
-### Out-of-Scope Use
-
-*List how the model may foreseeably be misused and address what users ought not to do with the model.*
--->
-
-<!--
-## Bias, Risks and Limitations
+---
 
-*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
--->
+## 📊 評価 / Evaluation
 
-<!--
-### Recommendations
+本モデルは、内部評価として MTEB (Massive Text Embedding Benchmark) 相当の環境でテストされ、  
+前バージョンよりもさらに高い精度を達成しています。
 
-*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
--->
+### **CodeSearchNetRetrieval (標準版) 成績**
 
-## Training Details
+| メトリクス         | スコア    |
+|--------------------|-----------|
+| **nDCG@10**         | **0.8946** |
+| Recall@10          | 0.9597    |
+| MAP@10             | 0.8731    |
+| MRR@10             | 0.8731    |
 
-### Training Dataset
+詳細スコア（抜粋）：
+- `ndcg_at_1`: 0.8175
+- `ndcg_at_3`: 0.8810
+- `ndcg_at_5`: 0.8888
+- `recall_at_20`: 0.9672
+- `recall_at_100`: 0.9807
+- `recall_at_1000`: 1.0
 
-#### Unnamed Dataset
+> 前バージョン (`Crow-Plus`) 比較で nDCG@10 がわずかに向上（+0.17pt）しており、より安定した高リコールが得られています。
 
-* Size: 2,022,217 training samples
-* Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
-* Approximate statistics based on the first 1000 samples:
-  |         | sentence_0                                                                          | sentence_1                                                                            | label                                                         |
-  |:--------|:------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------------|:--------------------------------------------------------------|
-  | type    | string                                                                              | string                                                                                | float                                                         |
-  | details | <ul><li>min: 3 tokens</li><li>mean: 48.56 tokens</li><li>max: 1024 tokens</li></ul> | <ul><li>min: 30 tokens</li><li>mean: 171.79 tokens</li><li>max: 1024 tokens</li></ul> | <ul><li>min: 1.0</li><li>mean: 1.0</li><li>max: 1.0</li></ul> |
-* Samples:
-  | sentence_0                                                                                | sentence_1                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      | label            |
-  |:------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:-----------------|
-  | <code>// GetNodeID returns the NodeID field if it's non-nil, zero value otherwise.</code> | <code>func (a *App) GetNodeID() string {<br>	if a == nil || a.NodeID == nil {<br>		return ""<br>	}<br>	return *a.NodeID<br>}</code>                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             | <code>1.0</code> |
-  | <code>// _NET_WM_STRUT_PARTIAL set</code>                                                 | <code>func WmStrutPartialSet(xu *xgbutil.XUtil, win xproto.Window,<br>	struts *WmStrutPartial) error {<br><br>	rawStruts := make([]uint, 12)<br>	rawStruts[0] = struts.Left<br>	rawStruts[1] = struts.Right<br>	rawStruts[2] = struts.Top<br>	rawStruts[3] = struts.Bottom<br>	rawStruts[4] = struts.LeftStartY<br>	rawStruts[5] = struts.LeftEndY<br>	rawStruts[6] = struts.RightStartY<br>	rawStruts[7] = struts.RightEndY<br>	rawStruts[8] = struts.TopStartX<br>	rawStruts[9] = struts.TopEndX<br>	rawStruts[10] = struts.BottomStartX<br>	rawStruts[11] = struts.BottomEndX<br><br>	return xprop.ChangeProp32(xu, win, "_NET_WM_STRUT_PARTIAL", "CARDINAL",<br>		rawStruts...)<br>}</code> | <code>1.0</code> |
-  | <code>//GetQyAccessToken 获取access_token</code>                                            | <code>func (ctx *Context) GetQyAccessToken() (accessToken string, err error) {<br>	ctx.accessTokenLock.Lock()<br>	defer ctx.accessTokenLock.Unlock()<br><br>	accessTokenCacheKey := fmt.Sprintf("qy_access_token_%s", ctx.AppID)<br>	val := ctx.Cache.Get(accessTokenCacheKey)<br>	if val != nil {<br>		accessToken = val.(string)<br>		return<br>	}<br><br>	//从微信服务器获取<br>	var resQyAccessToken ResQyAccessToken<br>	resQyAccessToken, err = ctx.GetQyAccessTokenFromServer()<br>	if err != nil {<br>		return<br>	}<br><br>	accessToken = resQyAccessToken.AccessToken<br>	return<br>}</code>                                                                                                  | <code>1.0</code> |
-* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
-  ```json
-  {
-      "scale": 20.0,
-      "similarity_fct": "cos_sim"
-  }
-  ```
+---
 
-### Training Hyperparameters
-#### Non-Default Hyperparameters
+### **COIRCodeSearchNetRetrieval 成績**
 
-- `per_device_train_batch_size`: 256
-- `per_device_eval_batch_size`: 256
-- `num_train_epochs`: 5
-- `fp16`: True
-- `multi_dataset_batch_sampler`: round_robin
+| メトリクス         | スコア    |
+|--------------------|-----------|
+| **nDCG@10**         | **0.8001** |
+| Recall@10          | 0.8806    |
+| MAP@10             | 0.7742    |
+| MRR@10             | 0.7742    |
 
-#### All Hyperparameters
-<details><summary>Click to expand</summary>
+詳細スコア（抜粋）：
+- `ndcg_at_1`: 0.7168
+- `ndcg_at_3`: 0.7775
+- `ndcg_at_5`: 0.7896
+- `recall_at_20`: 0.9057
+- `recall_at_100`: 0.9495
+- `recall_at_1000`: 0.9782
 
-- `overwrite_output_dir`: False
-- `do_predict`: False
-- `eval_strategy`: no
-- `prediction_loss_only`: True
-- `per_device_train_batch_size`: 256
-- `per_device_eval_batch_size`: 256
-- `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`: 5e-05
-- `weight_decay`: 0.0
-- `adam_beta1`: 0.9
-- `adam_beta2`: 0.999
-- `adam_epsilon`: 1e-08
-- `max_grad_norm`: 1
-- `num_train_epochs`: 5
-- `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
-- `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}
-- `tp_size`: 0
-- `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
-- `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`: batch_sampler
-- `multi_dataset_batch_sampler`: round_robin
+> こちらも、従来版より全体的に Recall 向上がみられ、特に Top-20、Top-100段階での検索精度に安定性が出ています。
 
-</details>
+---
 
-### Training Logs
-| Epoch  | Step  | Training Loss |
-|:------:|:-----:|:-------------:|
-| 0.0633 | 500   | 0.8015        |
-| 0.1266 | 1000  | 0.1036        |
-| 0.1899 | 1500  | 0.0973        |
-| 0.2532 | 2000  | 0.0921        |
-| 0.3165 | 2500  | 0.0876        |
-| 0.3797 | 3000  | 0.0861        |
-| 0.4430 | 3500  | 0.0843        |
-| 0.5063 | 4000  | 0.0841        |
-| 0.5696 | 4500  | 0.0788        |
-| 0.6329 | 5000  | 0.0794        |
-| 0.6962 | 5500  | 0.0782        |
-| 0.7595 | 6000  | 0.077         |
-| 0.8228 | 6500  | 0.0749        |
-| 0.8861 | 7000  | 0.0749        |
-| 0.9494 | 7500  | 0.0724        |
-| 1.0127 | 8000  | 0.0658        |
-| 1.0759 | 8500  | 0.0385        |
-| 1.1392 | 9000  | 0.0381        |
-| 1.2025 | 9500  | 0.0383        |
-| 1.2658 | 10000 | 0.0381        |
-| 1.3291 | 10500 | 0.0382        |
-| 1.3924 | 11000 | 0.0384        |
-| 1.4557 | 11500 | 0.0384        |
-| 1.5190 | 12000 | 0.039         |
-| 1.5823 | 12500 | 0.0391        |
-| 1.6456 | 13000 | 0.0401        |
-| 1.7089 | 13500 | 0.0383        |
-| 1.7722 | 14000 | 0.0392        |
-| 1.8354 | 14500 | 0.0371        |
-| 1.8987 | 15000 | 0.0387        |
-| 1.9620 | 15500 | 0.0385        |
-| 2.0253 | 16000 | 0.0298        |
-| 2.0886 | 16500 | 0.0171        |
-| 2.1519 | 17000 | 0.0174        |
-| 2.2152 | 17500 | 0.0171        |
-| 2.2785 | 18000 | 0.0169        |
-| 2.3418 | 18500 | 0.0174        |
-| 2.4051 | 19000 | 0.0177        |
-| 2.4684 | 19500 | 0.0175        |
-| 2.5316 | 20000 | 0.0171        |
-| 2.5949 | 20500 | 0.017         |
-| 2.6582 | 21000 | 0.0172        |
-| 2.7215 | 21500 | 0.0178        |
-| 2.7848 | 22000 | 0.0167        |
-| 2.8481 | 22500 | 0.0176        |
-| 2.9114 | 23000 | 0.0175        |
-| 2.9747 | 23500 | 0.0178        |
-| 3.0380 | 24000 | 0.0129        |
-| 3.1013 | 24500 | 0.0099        |
-| 3.1646 | 25000 | 0.0097        |
-| 3.2278 | 25500 | 0.0097        |
-| 3.2911 | 26000 | 0.0101        |
-| 3.3544 | 26500 | 0.0098        |
-| 3.4177 | 27000 | 0.0099        |
-| 3.4810 | 27500 | 0.0096        |
-| 3.5443 | 28000 | 0.0095        |
-| 3.6076 | 28500 | 0.0094        |
-| 3.6709 | 29000 | 0.0097        |
-| 3.7342 | 29500 | 0.01          |
-| 3.7975 | 30000 | 0.0096        |
-| 3.8608 | 30500 | 0.0098        |
-| 3.9241 | 31000 | 0.0095        |
-| 3.9873 | 31500 | 0.0094        |
-| 4.0506 | 32000 | 0.0079        |
-| 4.1139 | 32500 | 0.0074        |
-| 4.1772 | 33000 | 0.0072        |
-| 4.2405 | 33500 | 0.0073        |
-| 4.3038 | 34000 | 0.0071        |
-| 4.3671 | 34500 | 0.0073        |
-| 4.4304 | 35000 | 0.007         |
-| 4.4937 | 35500 | 0.0072        |
-| 4.5570 | 36000 | 0.0071        |
-| 4.6203 | 36500 | 0.0071        |
-| 4.6835 | 37000 | 0.0072        |
-| 4.7468 | 37500 | 0.0072        |
-| 4.8101 | 38000 | 0.0069        |
-| 4.8734 | 38500 | 0.007         |
-| 4.9367 | 39000 | 0.007         |
-| 5.0    | 39500 | 0.007         |
+## モデル変更点 / Improvements
 
+| 項目 | Crow-Plus | Crow-Plus-1.0 |
+|:---|:---|:---|
+| 学習データ | 従来データセット | PHP・Goのクリーン版追加 |
+| COIR nDCG@10 | 0.7988 | 0.8001 (+0.13pt) |
+| CodeSearchNet nDCG@10 | 0.8930 | 0.8946 (+0.16pt) |
+| Recall@10 | 0.9610 | 0.9597 (ほぼ同等) |
+| 特徴 | 標準構成 | データ品質向上による安定化 |
 
-### Framework Versions
-- Python: 3.11.11
-- Sentence Transformers: 3.4.1
-- Transformers: 4.51.3
-- PyTorch: 2.5.1+cu124
-- Accelerate: 1.3.0
-- Datasets: 3.5.0
-- Tokenizers: 0.21.0
+- MTEB公式には提出していないため、**非公式記録**です。
 
-## Citation
 
-### BibTeX
+## 使い方 / How to Use
 
-#### 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",
-}
-```
+従来と同様に、`sentence-transformers`ライブラリを用いて簡単に利用可能です。
+（※詳しい使用例はベース版に準じますので省略）
 
-#### MultipleNegativesRankingLoss
-```bibtex
-@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}
-}
+```python
+from sentence_transformers import SentenceTransformer
+model = SentenceTransformer("Shuu12121/CodeSearch-ModernBERT-Crow-Plus-1.0")
 ```
 
-<!--
-## Glossary
+---
 
-*Clearly define terms in order to be accessible across audiences.*
--->
+## 注意事項 / Limitations
 
-<!--
-## Model Card Authors
+- 関数レベルのコード検索に最適化されており、巨大なファイルや不完全コードには注意が必要です。
+- クリーン版追加による改善はあくまで「一般傾向」であり、特定ドメインにおいては追加チューニングが効果的な場合もあります。
+- 本バージョンの結果はMTEB公式には提出されていないため、**あくまで参考記録**である点にご留意ください。
 
-*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
--->
+---
 
-<!--
-## Model Card Contact
+## 連絡先 / Contact
 
-*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
--->
+ご質問・フィードバックは、開発者 [Shuu12121](https://huggingface.co/Shuu12121) までご連絡ください。  
+📧 shun0212114@outlook.jp