{ "schema_version": "2", "name": "Inkling VQA", "runtime": { "default": "client" }, "references": [ { "id": "ref_image", "role": "reference", "label": "Image", "asset_type": "image", "inputs": [{ "id": "in", "label": "Image", "type": "image" }], "outputs": [{ "id": "out", "label": "Image", "type": "image" }], "x": 80, "y": 150, "width": 220, "height": 122, "data": {} }, { "id": "ref_question", "role": "reference", "label": "Question", "asset_type": "text", "inputs": [{ "id": "in", "label": "Text", "type": "text" }], "outputs": [{ "id": "out", "label": "Text", "type": "text" }], "x": 80, "y": 320, "width": 220, "height": 161, "data": { "in": "What is in this image?" } } ], "operators": [ { "id": "model_inkling", "role": "operator", "kind": "model", "label": "Inkling", "model_id": "thinkingmachines/Inkling", "pipeline_tag": "image-text-to-text", "endpoint": "visual_question_answering", "inputs": [ { "id": "image", "label": "Image", "type": "image" }, { "id": "question", "label": "Question", "type": "text" } ], "outputs": [ { "id": "out_0", "label": "Answer", "type": "text", "output_index": 0 } ], "x": 380, "y": 200, "width": 220, "height": 130, "data": {} } ], "subjects": [ { "id": "subj_answer", "role": "subject", "label": "Answer", "asset_type": "text", "inputs": [{ "id": "in", "label": "Text", "type": "text" }], "outputs": [{ "id": "out", "label": "Text", "type": "text" }], "x": 680, "y": 230, "width": 220, "height": 105, "data": {} } ], "edges": [ { "id": "e1", "from_node_id": "ref_image", "from_port_id": "out", "to_node_id": "model_inkling", "to_port_id": "image", "type": "image" }, { "id": "e2", "from_node_id": "ref_question", "from_port_id": "out", "to_node_id": "model_inkling", "to_port_id": "question", "type": "text" }, { "id": "e3", "from_node_id": "model_inkling", "from_port_id": "out_0", "to_node_id": "subj_answer", "to_port_id": "in", "type": "text" } ], "view": { "default": "canvas" } }