| { |
| "architecture": { |
| "family": "image_classification", |
| "is_moe": false, |
| "view": "encoder" |
| }, |
| "capabilities": { |
| "task_heads": [ |
| "backbone", |
| "image", |
| "image_classification" |
| ], |
| "tensor_parallel": false |
| }, |
| "components": [ |
| { |
| "children": [ |
| "embeddings", |
| "encoder", |
| "layernorm" |
| ], |
| "class_name": "ConvNextV2Model", |
| "id": "model", |
| "kind": "model", |
| "path_pattern": "model" |
| }, |
| { |
| "attributes": {}, |
| "children": [ |
| "embeddings.patch_embeddings", |
| "embeddings.layernorm" |
| ], |
| "class_name": "ConvNextV2Embeddings", |
| "id": "embeddings", |
| "kind": "embedding", |
| "path_pattern": "model.embeddings" |
| }, |
| { |
| "class_name": "Conv2d", |
| "id": "embeddings.patch_embeddings", |
| "kind": "convolution", |
| "path_pattern": "model.embeddings.patch_embeddings" |
| }, |
| { |
| "attributes": { |
| "norm_type": "layer" |
| }, |
| "class_name": "ConvNextV2LayerNorm", |
| "id": "embeddings.layernorm", |
| "kind": "normalization", |
| "path_pattern": "model.embeddings.layernorm" |
| }, |
| { |
| "children": [ |
| "encoder_conv_next_v2_stages" |
| ], |
| "class_name": "ConvNextV2Encoder", |
| "id": "encoder", |
| "kind": "encoder", |
| "path_pattern": "model.encoder" |
| }, |
| { |
| "attributes": { |
| "norm_type": "layer" |
| }, |
| "class_name": "LayerNorm", |
| "id": "layernorm", |
| "kind": "normalization", |
| "path_pattern": "model.layernorm" |
| } |
| ], |
| "config": { |
| "class_name": "ConvNextV2Config", |
| "model_type": "convnextv2", |
| "module": "transformers.models.convnextv2.configuration_convnextv2", |
| "referenced_fields": { |
| "depths": [ |
| 3, |
| 3, |
| 9, |
| 3 |
| ] |
| }, |
| "salient_fields": { |
| "hidden_act": "gelu", |
| "is_encoder_decoder": false |
| } |
| }, |
| "dataflow": { |
| "input": { |
| "name": "pixel_values", |
| "shape": [ |
| "B", |
| 3, |
| 224, |
| 224 |
| ] |
| }, |
| "output": { |
| "shape": [ |
| "B", |
| 768, |
| 7, |
| 7 |
| ] |
| }, |
| "shapes": { |
| "embeddings": { |
| "in": [ |
| "B", |
| 3, |
| 224, |
| 224 |
| ], |
| "out": [ |
| "B", |
| 96, |
| 56, |
| 56 |
| ] |
| }, |
| "encoder": { |
| "in": [ |
| "B", |
| 96, |
| 56, |
| 56 |
| ], |
| "out": [ |
| "B", |
| 768, |
| 7, |
| 7 |
| ] |
| }, |
| "layernorm": { |
| "in": [ |
| "B", |
| 768 |
| ], |
| "out": [ |
| "B", |
| 768 |
| ] |
| } |
| }, |
| "source": "observed_forward_meta" |
| }, |
| "edges": [ |
| { |
| "kind": "data", |
| "source": "embeddings", |
| "target": "encoder" |
| }, |
| { |
| "kind": "data", |
| "source": "encoder", |
| "target": "layernorm" |
| } |
| ], |
| "extends": null, |
| "model_type": "convnextv2", |
| "provenance": { |
| "config_class": "ConvNextV2Config", |
| "config_module": "transformers.models.convnextv2.configuration_convnextv2", |
| "model_class": "ConvNextV2Model", |
| "model_module": "transformers.models.convnextv2.modeling_convnextv2" |
| }, |
| "repeats": [ |
| { |
| "body": "encoder_conv_next_v2_stage", |
| "container_path_pattern": "model.encoder.stages", |
| "count": 4, |
| "count_expr": "4", |
| "count_source": "module_tree", |
| "id": "encoder_conv_next_v2_stages", |
| "index_symbol": "i", |
| "item_path_pattern": "model.encoder.stages.{i}", |
| "kind": "symbolic_repeat", |
| "provenance": { |
| "class_name": "ConvNextV2Stage", |
| "container_path_pattern": "model.encoder.stages", |
| "item_path_pattern": "model.encoder.stages.{i}", |
| "source": "module_tree_repeat_collapse" |
| }, |
| "repeated_class_name": "ConvNextV2Stage" |
| }, |
| { |
| "body": "encoder_layer", |
| "container_path_pattern": "model.encoder.stages.{i}.layers", |
| "count": 3, |
| "count_expr": "config.depths[i]", |
| "count_source": "config", |
| "id": "encoder_layers", |
| "index_symbol": "j", |
| "item_path_pattern": "model.encoder.stages.{i}.layers.{j}", |
| "kind": "symbolic_repeat", |
| "provenance": { |
| "class_name": "ConvNextV2Layer", |
| "container_path_pattern": "model.encoder.stages.{i}.layers", |
| "item_path_pattern": "model.encoder.stages.{i}.layers.{j}", |
| "source": "module_tree_repeat_collapse" |
| }, |
| "repeated_class_name": "ConvNextV2Layer" |
| } |
| ], |
| "schema_version": "architecture-template-v0", |
| "templates": [ |
| { |
| "children": [ |
| "encoder_conv_next_v2_stage.downsampling_layer", |
| "encoder_layers" |
| ], |
| "class_name": "ConvNextV2Stage", |
| "id": "encoder_conv_next_v2_stage", |
| "kind": "module", |
| "path_pattern": "model.encoder.stages.{i}" |
| }, |
| { |
| "class_name": "ModuleList", |
| "id": "encoder_conv_next_v2_stage.downsampling_layer", |
| "kind": "repeated_container", |
| "path_pattern": "model.encoder.stages.{i}.downsampling_layer" |
| }, |
| { |
| "children": [ |
| "encoder_layer.dwconv", |
| "encoder_layer.layernorm" |
| ], |
| "class_name": "ConvNextV2Layer", |
| "id": "encoder_layer", |
| "kind": "convolution", |
| "path_pattern": "model.encoder.stages.{i}.layers.{j}" |
| }, |
| { |
| "class_name": "Conv2d", |
| "id": "encoder_layer.dwconv", |
| "kind": "convolution", |
| "path_pattern": "model.encoder.stages.{i}.layers.{j}.dwconv" |
| }, |
| { |
| "attributes": { |
| "norm_type": "layer" |
| }, |
| "class_name": "ConvNextV2LayerNorm", |
| "id": "encoder_layer.layernorm", |
| "kind": "normalization", |
| "path_pattern": "model.encoder.stages.{i}.layers.{j}.layernorm" |
| } |
| ] |
| } |
|
|
