This is a Cross Encoder model finetuned from cross-encoder/ms-marco-MiniLM-L6-v2 using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

Model Type: Cross Encoder
Base model: cross-encoder/ms-marco-MiniLM-L6-v2
Maximum Sequence Length: 512 tokens
Number of Output Labels: 1 label
Supported Modality: Text

Model Sources

Documentation: Sentence Transformers Documentation
Documentation: Cross Encoder Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Cross Encoders on Hugging Face

Full Model Architecture

CrossEncoder(
  (0): Transformer({'transformer_task': 'sequence-classification', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'logits'}}, 'module_output_name': 'scores', 'architecture': 'BertForSequenceClassification'})
)

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("cross_encoder_model_id")
# Get scores for pairs of inputs
pairs = [
    ['Tool name: get_items_for_portfolio\nTool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.\nTool input parameters:\n- limit: number (optional) — Results per page (1-100)\n- offset: string (optional) — Pagination offset token\n- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response\n- portfolio_gid: string (required) — Globally unique identifier for the portfolio', 'API operationId: getItemsForPortfolio\nAPI method: GET\nAPI path: /portfolios/{portfolio_gid}/items\nAPI summary: Get portfolio items\nAPI description: <b>Required scope: </b><code>portfolios:read</code>\n\nGet a list of the items in compact form in a portfolio.\nAPI parameters:\n- ref=portfolio_path_gid\n- ref=pretty\n- ref=limit\n- ref=offset\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["portfolios:read"]}]'],
    ['Tool name: get_items_for_portfolio\nTool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.\nTool input parameters:\n- limit: number (optional) — Results per page (1-100)\n- offset: string (optional) — Pagination offset token\n- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response\n- portfolio_gid: string (required) — Globally unique identifier for the portfolio', 'API operationId: getGoals\nAPI method: GET\nAPI path: /goals\nAPI summary: Get goals\nAPI description: <b>Required scope: </b><code>goals:read</code>\n\nReturns compact goal records.\nAPI parameters:\n- ref=pretty\n- portfolio in=query type=string optional — Globally unique identifier for supporting portfolio.\n- project in=query type=string optional — Globally unique identifier for supporting project.\n- task in=query type=string optional — Globally unique identifier for supporting task.\n- is_workspace_level in=query type=boolean optional — Filter to goals with is_workspace_level set to query value. Must be used with the workspace parameter.\n- team in=query type=string optional — Globally unique identifier for the team.\n- workspace in=query type=string optional — Globally unique identifier for the workspace.\n- time_periods in=query type=array optional — Globally unique identifiers for the time periods.\n- ref=limit\n- ref=offset\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["goals:read"]}]'],
    ['Tool name: get_items_for_portfolio\nTool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.\nTool input parameters:\n- limit: number (optional) — Results per page (1-100)\n- offset: string (optional) — Pagination offset token\n- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response\n- portfolio_gid: string (required) — Globally unique identifier for the portfolio', 'API operationId: searchProjectsForWorkspace\nAPI method: GET\nAPI path: /workspaces/{workspace_gid}/projects/search\nAPI summary: Search projects in a workspace\nAPI description: <b>Required scope: </b><code>projects:read</code>\n\nTo mirror the functionality of the Asana web app\'s advanced search feature, the Asana API has a project search endpoint that allows you to build complex filters to find and retrieve the exact data you need.\n#### Premium access\nLike the Asana web product\'s advance search feature, this search endpoint will only be available to premium Asana users. A user is premium if any of the following is true:\n\n- The workspace in which the search is being performed is a premium workspace\n- The user is a member of a premium team inside the workspace\n\nEven if a user is only a member of a premium team inside a non-premium workspace, search will allow them to find data anywhere in the workspace, not just inside the premium team. Making a search request using credentials of a non-premium user will result in a `402 Payment Required` error.\n#### Pagination\nSearch results are not stable; repeating the same query multiple times may return the data in a different order, even if the data do not change. Because of this, the traditional [pagination](/docs/pagination) available elsewhere in the Asana API is not available here. However, you can paginate manually by sorting the search results by their creation time and then modifying each subsequent query to exclude data you have already seen. Page sizes are limited to a maximum of 100 items, and can be specified by the `limit` query parameter.\n#### Eventual consistency\nChanges in Asana (regardless of whether they’re made though the web product or the API) are forwarded to our search infrastructure to be indexed. This process can take between 10 and 60 seconds to complete under normal operation, and longer during some production incidents. Making a change to a project that would alter its presence in a particular search query will not be reflected immediately. This is also true of the advanced search feature in the web product.\nBecause of this delay, the search endpoint is not suited for use cases that require immediate consistency after writes. If you need read-your-write behavior or strongly consistent results, we recommend using [Get multiple projects](/reference/getprojects) instead.\n#### Rate limits\nYou may receive a `429 Too Many Requests` response if you hit any of our [rate limits](/docs/rate-limits).\n#### Custom field parameters\n| Parameter name | Custom field type | Accepted type |\n|---|---|---|\n| custom_fields.{gid}.is_set | All | Boolean |\n| custom_fields.{gid}.value | Text | String |\n| custom_fields.{gid}.value | Number | Number |\n| custom_fields.{gid}.value | Enum | Enum option ID |\n| custom_fields.{gid}.starts_with | Text only | String |\n| custom_fields.{gid}.ends_with | Text only | String |\n| custom_fields.{gid}.contains | Text only | String |\n| custom_fields.{gid}.less_than | Number only | Number |\n| custom_fields.{gid}.greater_than | Number only | Number |\n\n\nFor example, if the gid of the custom field is 12345, the query parameter to find projects where it i…\nAPI parameters:\n- ref=workspace_path_gid\n- ref=pretty\n- text in=query type=string optional — Performs full-text search on the project name.\n- sort_by in=query type=string optional — One of `due_date`, `created_at`, `completed_at`, `relevance`, or `modified_at`, defaults to `modified_at`.\n- sort_ascending in=query type=boolean optional — Default `false`.\n- completed in=query type=boolean optional — Filter on project completion status.\n- teams.any in=query type=string optional — Comma-separated list of team IDs.\n- owner.any in=query type=string optional — Comma-separated list of user identifiers to filter on as project owners. This can either be the string "me", an email, or the gid of a user.\n- members.any in=query type=string optional — Comma-separated list of user identifiers to filter on as members. This can either be the string "me", an email, or the gid of a user.\n- members.not in=query type=string optional — Comma-separated list of user identifiers to exclude as members. This can either be the string "me", an email, or the gid of a user.\n- portfolios.any in=query type=string optional — Comma-separated list of portfolio IDs to filter on.\n- completed_on in=query type=string optional — ISO 8601 date string or `null`.\n- completed_on.before in=query type=string optional — ISO 8601 date string.\n- completed_on.after in=query type=string optional — ISO 8601 date string.\n- completed_at.before in=query type=string optional — ISO 8601 datetime string.\n- completed_at.after in=query type=string optional — ISO 8601 datetime string.\n- created_on in=query type=string optional — ISO 8601 date string or `null`.\n- created_on.before in=query type=string optional — ISO 8601 date string.\n- created_on.after in=query type=string optional — ISO 8601 date string.\n- created_at.before in=query type=string optional — ISO 8601 datetime string.\n- created_at.after in=query type=string optional — ISO 8601 datetime string.\n- due_on in=query type=string optional — ISO 8601 date string or `null`.\n- due_on.before in=query type=string optional — ISO 8601 date string.\n- due_on.after in=query type=string optional — ISO 8601 date string.\n- due_at.before in=query type=string optional — ISO 8601 datetime string.\n- due_at.after in=query type=string optional — ISO 8601 datetime string.\n- start_on in=query type=string optional — ISO 8601 date string or `null`.\n- start_on.before in=query type=string optional — ISO 8601 date string.\n- start_on.after in=query type=string optional — ISO 8601 date string.\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["projects:read"]}]'],
    ['Tool name: get_items_for_portfolio\nTool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.\nTool input parameters:\n- limit: number (optional) — Results per page (1-100)\n- offset: string (optional) — Pagination offset token\n- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response\n- portfolio_gid: string (required) — Globally unique identifier for the portfolio', 'API operationId: removeDependentsForTask\nAPI method: POST\nAPI path: /tasks/{task_gid}/removeDependents\nAPI summary: Unlink dependents from a task\nAPI description: <b>Required scope: </b><code>tasks:write</code>\n\nUnlinks a set of dependents from this task.\nAPI parameters:\n- ref=task_path_gid\n- ref=pretty\nAPI request body:\nrequired\napplication/json: object\n- data: ModifyDependentsRequest (optional)\nAPI security: [{"personalAccessToken":[]},{"oauth2":["tasks:write"]}]'],
    ['Tool name: create_project_status_update\nTool description: Post a status update to a project or portfolio. Use for project health updates, milestone documentation, or blocker reporting. Returns created status with gid, parent, title, status_type, author, created_at, permalink_url. Exactly one of text or html_text must be provided.\nTool input parameters:\n- text: string (optional) — Plain-text status body. Exactly one of text or html_text must be provided.\n- color: string (required) — Status color: green (on track), yellow (at risk), red (off track), blue (on hold), complete\n- title: string (required) — Status title\n- parent: string (required) — GID of the project or portfolio to post the status update to.\n- html_text: string (optional) — Exactly one of text or html_text must be provided. Allowed elements only: <body>, <strong>, <em>, <u>, <s>, <code>, <ol>, <ul>, <li>, <a>, <blockquote>, <pre>. Do not use <h1>, <h2>, <hr/>, or <img> here. Must be well-formed XML with a sin…', 'API operationId: createProjectStatusForProject\nAPI method: POST\nAPI path: /projects/{project_gid}/project_statuses\nAPI summary: Create a project status\nAPI description: *Deprecated: new integrations should prefer the `/status_updates` route.*\n\nCreates a new status update on the project.\n\nReturns the full record of the newly created project status update.\nAPI parameters:\n- ref=project_path_gid\n- ref=pretty\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nrequired\napplication/json: object\n- data: ProjectStatusRequest (optional)\nAPI security: [{"personalAccessToken":[]},{"oauth2":[]}]'],
]
scores = model.predict(pairs)
print(scores)
# [0.9791 0.5527 0.0071 0.001  0.9961]

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'Tool name: get_items_for_portfolio\nTool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.\nTool input parameters:\n- limit: number (optional) — Results per page (1-100)\n- offset: string (optional) — Pagination offset token\n- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response\n- portfolio_gid: string (required) — Globally unique identifier for the portfolio',
    [
        'API operationId: getItemsForPortfolio\nAPI method: GET\nAPI path: /portfolios/{portfolio_gid}/items\nAPI summary: Get portfolio items\nAPI description: <b>Required scope: </b><code>portfolios:read</code>\n\nGet a list of the items in compact form in a portfolio.\nAPI parameters:\n- ref=portfolio_path_gid\n- ref=pretty\n- ref=limit\n- ref=offset\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["portfolios:read"]}]',
        'API operationId: getGoals\nAPI method: GET\nAPI path: /goals\nAPI summary: Get goals\nAPI description: <b>Required scope: </b><code>goals:read</code>\n\nReturns compact goal records.\nAPI parameters:\n- ref=pretty\n- portfolio in=query type=string optional — Globally unique identifier for supporting portfolio.\n- project in=query type=string optional — Globally unique identifier for supporting project.\n- task in=query type=string optional — Globally unique identifier for supporting task.\n- is_workspace_level in=query type=boolean optional — Filter to goals with is_workspace_level set to query value. Must be used with the workspace parameter.\n- team in=query type=string optional — Globally unique identifier for the team.\n- workspace in=query type=string optional — Globally unique identifier for the workspace.\n- time_periods in=query type=array optional — Globally unique identifiers for the time periods.\n- ref=limit\n- ref=offset\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["goals:read"]}]',
        'API operationId: searchProjectsForWorkspace\nAPI method: GET\nAPI path: /workspaces/{workspace_gid}/projects/search\nAPI summary: Search projects in a workspace\nAPI description: <b>Required scope: </b><code>projects:read</code>\n\nTo mirror the functionality of the Asana web app\'s advanced search feature, the Asana API has a project search endpoint that allows you to build complex filters to find and retrieve the exact data you need.\n#### Premium access\nLike the Asana web product\'s advance search feature, this search endpoint will only be available to premium Asana users. A user is premium if any of the following is true:\n\n- The workspace in which the search is being performed is a premium workspace\n- The user is a member of a premium team inside the workspace\n\nEven if a user is only a member of a premium team inside a non-premium workspace, search will allow them to find data anywhere in the workspace, not just inside the premium team. Making a search request using credentials of a non-premium user will result in a `402 Payment Required` error.\n#### Pagination\nSearch results are not stable; repeating the same query multiple times may return the data in a different order, even if the data do not change. Because of this, the traditional [pagination](/docs/pagination) available elsewhere in the Asana API is not available here. However, you can paginate manually by sorting the search results by their creation time and then modifying each subsequent query to exclude data you have already seen. Page sizes are limited to a maximum of 100 items, and can be specified by the `limit` query parameter.\n#### Eventual consistency\nChanges in Asana (regardless of whether they’re made though the web product or the API) are forwarded to our search infrastructure to be indexed. This process can take between 10 and 60 seconds to complete under normal operation, and longer during some production incidents. Making a change to a project that would alter its presence in a particular search query will not be reflected immediately. This is also true of the advanced search feature in the web product.\nBecause of this delay, the search endpoint is not suited for use cases that require immediate consistency after writes. If you need read-your-write behavior or strongly consistent results, we recommend using [Get multiple projects](/reference/getprojects) instead.\n#### Rate limits\nYou may receive a `429 Too Many Requests` response if you hit any of our [rate limits](/docs/rate-limits).\n#### Custom field parameters\n| Parameter name | Custom field type | Accepted type |\n|---|---|---|\n| custom_fields.{gid}.is_set | All | Boolean |\n| custom_fields.{gid}.value | Text | String |\n| custom_fields.{gid}.value | Number | Number |\n| custom_fields.{gid}.value | Enum | Enum option ID |\n| custom_fields.{gid}.starts_with | Text only | String |\n| custom_fields.{gid}.ends_with | Text only | String |\n| custom_fields.{gid}.contains | Text only | String |\n| custom_fields.{gid}.less_than | Number only | Number |\n| custom_fields.{gid}.greater_than | Number only | Number |\n\n\nFor example, if the gid of the custom field is 12345, the query parameter to find projects where it i…\nAPI parameters:\n- ref=workspace_path_gid\n- ref=pretty\n- text in=query type=string optional — Performs full-text search on the project name.\n- sort_by in=query type=string optional — One of `due_date`, `created_at`, `completed_at`, `relevance`, or `modified_at`, defaults to `modified_at`.\n- sort_ascending in=query type=boolean optional — Default `false`.\n- completed in=query type=boolean optional — Filter on project completion status.\n- teams.any in=query type=string optional — Comma-separated list of team IDs.\n- owner.any in=query type=string optional — Comma-separated list of user identifiers to filter on as project owners. This can either be the string "me", an email, or the gid of a user.\n- members.any in=query type=string optional — Comma-separated list of user identifiers to filter on as members. This can either be the string "me", an email, or the gid of a user.\n- members.not in=query type=string optional — Comma-separated list of user identifiers to exclude as members. This can either be the string "me", an email, or the gid of a user.\n- portfolios.any in=query type=string optional — Comma-separated list of portfolio IDs to filter on.\n- completed_on in=query type=string optional — ISO 8601 date string or `null`.\n- completed_on.before in=query type=string optional — ISO 8601 date string.\n- completed_on.after in=query type=string optional — ISO 8601 date string.\n- completed_at.before in=query type=string optional — ISO 8601 datetime string.\n- completed_at.after in=query type=string optional — ISO 8601 datetime string.\n- created_on in=query type=string optional — ISO 8601 date string or `null`.\n- created_on.before in=query type=string optional — ISO 8601 date string.\n- created_on.after in=query type=string optional — ISO 8601 date string.\n- created_at.before in=query type=string optional — ISO 8601 datetime string.\n- created_at.after in=query type=string optional — ISO 8601 datetime string.\n- due_on in=query type=string optional — ISO 8601 date string or `null`.\n- due_on.before in=query type=string optional — ISO 8601 date string.\n- due_on.after in=query type=string optional — ISO 8601 date string.\n- due_at.before in=query type=string optional — ISO 8601 datetime string.\n- due_at.after in=query type=string optional — ISO 8601 datetime string.\n- start_on in=query type=string optional — ISO 8601 date string or `null`.\n- start_on.before in=query type=string optional — ISO 8601 date string.\n- start_on.after in=query type=string optional — ISO 8601 date string.\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nnone\nAPI security: [{"personalAccessToken":[]},{"oauth2":["projects:read"]}]',
        'API operationId: removeDependentsForTask\nAPI method: POST\nAPI path: /tasks/{task_gid}/removeDependents\nAPI summary: Unlink dependents from a task\nAPI description: <b>Required scope: </b><code>tasks:write</code>\n\nUnlinks a set of dependents from this task.\nAPI parameters:\n- ref=task_path_gid\n- ref=pretty\nAPI request body:\nrequired\napplication/json: object\n- data: ModifyDependentsRequest (optional)\nAPI security: [{"personalAccessToken":[]},{"oauth2":["tasks:write"]}]',
        'API operationId: createProjectStatusForProject\nAPI method: POST\nAPI path: /projects/{project_gid}/project_statuses\nAPI summary: Create a project status\nAPI description: *Deprecated: new integrations should prefer the `/status_updates` route.*\n\nCreates a new status update on the project.\n\nReturns the full record of the newly created project status update.\nAPI parameters:\n- ref=project_path_gid\n- ref=pretty\n- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…\nAPI request body:\nrequired\napplication/json: object\n- data: ProjectStatusRequest (optional)\nAPI security: [{"personalAccessToken":[]},{"oauth2":[]}]',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Training Details

Training Dataset

Unnamed Dataset

Size: 1,162 training samples
Columns: tool_text, api_text, and label
Approximate statistics based on the first 100 samples:
tool_text api_text label
type string string float
modality text text
details
min: 86 tokens
mean: 257.85 tokens
max: 512 tokens
min: 110 tokens
mean: 359.87 tokens
max: 512 tokens
min: 0.0
mean: 0.25
max: 1.0

Samples:

tool_text	api_text	label
Tool name: get_workspace_agents[object Object]Tool description: Returns a list of AI Teammate agents (automated agents, not human users) configured in a workspace. AI Teammates are Asana-specific automation agents — they are distinct from human coworkers, teammates, or workspace members. Do NOT use this tool when the user asks about people, users, teammates, or team members; use get_users or search_objects instead. Each agent record includes GID, name, resource_type, and resource_subtype. Use to discover available AI Teammate agents before fetching full details with get_agent. Supports pagination for workspaces with many agents.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100).[object Object]- offset: string (optional) — Pagination offset token from a previous response.[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response. Available fields: gid, resource_type, resource_subtype, name.	API operationId: getAgentsForWorkspace[object Object]API method: GET[object Object]API path: /workspaces/{workspace_gid}/agents[object Object]API summary: Get a list of agents in a workspace[object Object]API description: Returns the compact records of agents (AI Teammates) configured. Use [object Object] to request additional fields in the workspace.[object Object]API parameters:[object Object]- ref=workspace_path_gid[object Object]- ref=pretty[object Object]- ref=limit[object Object]- ref=offset[object Object]- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…[object Object]API request body:[object Object]none[object Object]API security: [{"personalAccessToken":[]},{"oauth2":[]}]	`1.0`
Tool name: get_workspace_agents[object Object]Tool description: Returns a list of AI Teammate agents (automated agents, not human users) configured in a workspace. AI Teammates are Asana-specific automation agents — they are distinct from human coworkers, teammates, or workspace members. Do NOT use this tool when the user asks about people, users, teammates, or team members; use get_users or search_objects instead. Each agent record includes GID, name, resource_type, and resource_subtype. Use to discover available AI Teammate agents before fetching full details with get_agent. Supports pagination for workspaces with many agents.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100).[object Object]- offset: string (optional) — Pagination offset token from a previous response.[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response. Available fields: gid, resource_type, resource_subtype, name.	API operationId: typeaheadForWorkspace[object Object]API method: GET[object Object]API path: /workspaces/{workspace_gid}/typeahead[object Object]API summary: Get objects via typeahead[object Object]API description: [object Object][object Object][object Object][object Object]Retrieves objects in the workspace based via an auto-completion/typeahead[object Object]search algorithm. This feature is meant to provide results quickly, so do[object Object]not rely on this API to provide extremely accurate search results. The[object Object]result set is limited to a single page of results with a maximum size, so[object Object]you won’t be able to fetch large numbers of results.[object Object][object Object]The typeahead search API provides search for objects from a single[object Object]workspace. This endpoint should be used to query for objects when[object Object]creating an auto-completion/typeahead search feature. This API is meant[object Object]to provide results quickly and should not be relied upon for accurate or[object Object]exhaustive search results. The results sets are limited in size and[object Object]cannot be paginated.[object Object][object Object]Queries return a compact representation of each object which is...	`0.0`
Tool name: get_workspace_agents[object Object]Tool description: Returns a list of AI Teammate agents (automated agents, not human users) configured in a workspace. AI Teammates are Asana-specific automation agents — they are distinct from human coworkers, teammates, or workspace members. Do NOT use this tool when the user asks about people, users, teammates, or team members; use get_users or search_objects instead. Each agent record includes GID, name, resource_type, and resource_subtype. Use to discover available AI Teammate agents before fetching full details with get_agent. Supports pagination for workspaces with many agents.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100).[object Object]- offset: string (optional) — Pagination offset token from a previous response.[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response. Available fields: gid, resource_type, resource_subtype, name.	API operationId: searchProjectsForWorkspace[object Object]API method: GET[object Object]API path: /workspaces/{workspace_gid}/projects/search[object Object]API summary: Search projects in a workspace[object Object]API description: [object Object][object Object][object Object][object Object]To mirror the functionality of the Asana web app's advanced search feature, the Asana API has a project search endpoint that allows you to build complex filters to find and retrieve the exact data you need.[object Object]#### Premium access[object Object]Like the Asana web product's advance search feature, this search endpoint will only be available to premium Asana users. A user is premium if any of the following is true:[object Object][object Object]- The workspace in which the search is being performed is a premium workspace[object Object]- The user is a member of a premium team inside the workspace[object Object][object Object]Even if a user is only a member of a premium team inside a non-premium workspace, search will allow them to find data anywhere in the workspace, not just inside the premium team. Making a search request using credentials of a non-pre...	`0.0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fn": "torch.nn.modules.linear.Identity", 3 "pos_weight": 3.194945812225342 4}

Evaluation Dataset

Unnamed Dataset

Size: 288 evaluation samples
Columns: tool_text, api_text, and label
Approximate statistics based on the first 100 samples:
tool_text api_text label
type string string float
modality text text
details
min: 53 tokens
mean: 303.04 tokens
max: 512 tokens
min: 101 tokens
mean: 349.38 tokens
max: 512 tokens
min: 0.0
mean: 0.25
max: 1.0

Samples:

tool_text	api_text	label
Tool name: get_items_for_portfolio[object Object]Tool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100)[object Object]- offset: string (optional) — Pagination offset token[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response[object Object]- portfolio_gid: string (required) — Globally unique identifier for the portfolio	API operationId: getItemsForPortfolio[object Object]API method: GET[object Object]API path: /portfolios/{portfolio_gid}/items[object Object]API summary: Get portfolio items[object Object]API description: [object Object][object Object][object Object][object Object]Get a list of the items in compact form in a portfolio.[object Object]API parameters:[object Object]- ref=portfolio_path_gid[object Object]- ref=pretty[object Object]- ref=limit[object Object]- ref=offset[object Object]- opt_fields in=query type=array optional — This endpoint returns a resource which excludes some properties by default. To include those optional properties, set this query parameter to a comma-separated list of the properties you wish to incl…[object Object]API request body:[object Object]none[object Object]API security: [{"personalAccessToken":[]},{"oauth2":["portfolios:read"]}]	`1.0`
Tool name: get_items_for_portfolio[object Object]Tool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100)[object Object]- offset: string (optional) — Pagination offset token[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response[object Object]- portfolio_gid: string (required) — Globally unique identifier for the portfolio	API operationId: getGoals[object Object]API method: GET[object Object]API path: /goals[object Object]API summary: Get goals[object Object]API description: [object Object][object Object][object Object][object Object]Returns compact goal records.[object Object]API parameters:[object Object]- ref=pretty[object Object]- portfolio in=query type=string optional — Globally unique identifier for supporting portfolio.[object Object]- project in=query type=string optional — Globally unique identifier for supporting project.[object Object]- task in=query type=string optional — Globally unique identifier for supporting task.[object Object]- is_workspace_level in=query type=boolean optional — Filter to goals with is_workspace_level set to query value. Must be used with the workspace parameter.[object Object]- team in=query type=string optional — Globally unique identifier for the team.[object Object]- workspace in=query type=string optional — Globally unique identifier for the workspace.[object Object]- time_periods in=query type=array optional — Globally unique identifiers for the time periods.[object Object]- ref=limit[object Object]- ref=offset[object Object]- opt_fields in=query type=array optional — This endpoint returns a r...	`0.0`
Tool name: get_items_for_portfolio[object Object]Tool description: List projects, goals, and other items in a portfolio. Returns item names, IDs, and types. Use for portfolio content exploration and management. Supports pagination for portfolios with many items.[object Object]Tool input parameters:[object Object]- limit: number (optional) — Results per page (1-100)[object Object]- offset: string (optional) — Pagination offset token[object Object]- opt_fields: string (optional) — Comma-separated list of optional fields to include in the response[object Object]- portfolio_gid: string (required) — Globally unique identifier for the portfolio	API operationId: searchProjectsForWorkspace[object Object]API method: GET[object Object]API path: /workspaces/{workspace_gid}/projects/search[object Object]API summary: Search projects in a workspace[object Object]API description: [object Object][object Object][object Object][object Object]To mirror the functionality of the Asana web app's advanced search feature, the Asana API has a project search endpoint that allows you to build complex filters to find and retrieve the exact data you need.[object Object]#### Premium access[object Object]Like the Asana web product's advance search feature, this search endpoint will only be available to premium Asana users. A user is premium if any of the following is true:[object Object][object Object]- The workspace in which the search is being performed is a premium workspace[object Object]- The user is a member of a premium team inside the workspace[object Object][object Object]Even if a user is only a member of a premium team inside a non-premium workspace, search will allow them to find data anywhere in the workspace, not just inside the premium team. Making a search request using credentials of a non-pre...	`0.0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fn": "torch.nn.modules.linear.Identity", 3 "pos_weight": 3.194945812225342 4}

Training Hyperparameters

Non-Default Hyperparameters

learning_rate: 2e-05
warmup_steps: 0.1
load_best_model_at_end: True

All Hyperparameters

Click to expand

per_device_train_batch_size: 8
num_train_epochs: 3.0
max_steps: -1
learning_rate: 2e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0.1
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1.0
label_smoothing_factor: 0.0
bf16: False
fp16: False
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 8
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: True
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 42
data_seed: None
use_cpu: False
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
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
warmup_ratio: None
local_rank: -1
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.0685	10	0.7372	-
0.1370	20	1.2605	-
0.2055	30	1.3301	-
0.2740	40	1.1441	-
0.3425	50	1.3737	-
0.4110	60	1.0229	-
0.4795	70	1.2095	-
0.5479	80	0.8421	-
0.6164	90	0.9955	-
0.6849	100	1.2571	-
0.7534	110	0.8670	-
0.8219	120	0.9455	-
0.8904	130	0.8816	-
0.9589	140	0.6555	-
1.0	146	-	0.9174
1.0274	150	0.7276	-
1.0959	160	0.7638	-
1.1644	170	0.6309	-
1.2329	180	1.0922	-
1.3014	190	0.6035	-
1.3699	200	0.6395	-
1.4384	210	0.7156	-
1.5068	220	0.5402	-
1.5753	230	0.8845	-
1.6438	240	0.5796	-
1.7123	250	0.8343	-
1.7808	260	0.7073	-
1.8493	270	0.6420	-
1.9178	280	0.7564	-
1.9863	290	0.7804	-
2.0	292	-	0.8595
2.0548	300	0.5839	-
2.1233	310	0.6399	-
2.1918	320	0.3549	-
2.2603	330	0.8993	-
2.3288	340	0.5939	-
2.3973	350	0.7083	-
2.4658	360	0.4796	-
2.5342	370	0.6031	-
2.6027	380	0.5991	-
2.6712	390	0.6392	-
2.7397	400	0.4957	-
2.8082	410	0.4726	-
2.8767	420	0.5366	-
2.9452	430	0.7931	-
3.0	438	-	0.8235

The bold row denotes the saved checkpoint.

Training Time

Training: 50.0 minutes
Evaluation: 1.4 minutes
Total: 51.4 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.6.1
Transformers: 5.14.1
PyTorch: 2.13.0+cu130
Accelerate: 1.14.0
Datasets: 5.0.1
Tokenizers: 0.22.2

Additional Resources

Training and Finetuning Reranker Models with Sentence Transformers: the end-to-end guide for training or finetuning Cross Encoder (reranker) models.
Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video reranker models through the same API.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: training multimodal Cross Encoders.

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",
}

	tool_text	api_text	label
type	string	string	float
modality	text	text
details	min: 86 tokens mean: 257.85 tokens max: 512 tokens	min: 110 tokens mean: 359.87 tokens max: 512 tokens	min: 0.0 mean: 0.25 max: 1.0

Yoonseok55/tool-api-mapper-reranker

CrossEncoder based on cross-encoder/ms-marco-MiniLM-L6-v2

Model Details

Model Description

Model Sources

Full Model Architecture

Usage

Direct Usage (Sentence Transformers)

Training Details

Training Dataset

Unnamed Dataset

Evaluation Dataset

Unnamed Dataset

Training Hyperparameters

Non-Default Hyperparameters

All Hyperparameters

Training Logs

Training Time

Framework Versions

Additional Resources

Citation

BibTeX

Sentence Transformers

Yoonseok55/
tool-api-mapper-reranker