{ "nodes": { "afmoe": { "children": [], "depth": 1, "diff_size": 43, "extends": "llama", "is_modular": true, "parents": [ "llama", "qwen2_moe", "gpt_oss" ], "root": "llama" }, "aimv2": { "children": [], "depth": 1, "diff_size": 91, "extends": "siglip", "is_modular": true, "parents": [ "siglip", "clip", "llama" ], "root": "siglip" }, "altclip": { "children": [], "depth": 1, "diff_size": 59, "extends": "clip", "is_modular": true, "parents": [ "clip", "roberta", "chinese_clip" ], "root": "clip" }, "apertus": { "children": [], "depth": 1, "diff_size": 33, "extends": "llama", "is_modular": true, "parents": [ "llama", "nemotron" ], "root": "llama" }, "arcee": { "children": [], "depth": 1, "diff_size": 26, "extends": "llama", "is_modular": true, "parents": [ "llama", "nemotron" ], "root": "llama" }, "aria": { "children": [], "depth": 1, "diff_size": 137, "extends": "llama", "is_modular": true, "parents": [ "llama", "llava" ], "root": "llama" }, "audio_spectrogram_transformer": { "children": [], "depth": 1, "diff_size": 34, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "audioflamingo3": { "children": [ "glmasr", "musicflamingo" ], "depth": 2, "diff_size": 24, "extends": "voxtral", "is_modular": true, "parents": [ "voxtral", "qwen2_audio", "whisper" ], "root": "qwen2_audio" }, "aya_vision": { "children": [ "cohere2_vision" ], "depth": 1, "diff_size": 10, "extends": "llava", "is_modular": true, "parents": [ "llava" ], "root": "llava" }, "bamba": { "children": [], "depth": 1, "diff_size": 39, "extends": "llama", "is_modular": true, "parents": [ "llama", "jamba", "mamba2" ], "root": "llama" }, "bart": { "children": [ "biogpt", "plbart" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "bart" }, "beit": { "children": [], "depth": 1, "diff_size": 82, "extends": "vit", "is_modular": true, "parents": [ "vit", "resnet", "swin" ], "root": "vit" }, "bert": { "children": [ "ernie", "nomic_bert", "roberta", "xlm_roberta_xl" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "bert" }, "biogpt": { "children": [], "depth": 1, "diff_size": 39, "extends": "bart", "is_modular": true, "parents": [ "bart", "opt" ], "root": "bart" }, "bitnet": { "children": [], "depth": 1, "diff_size": 8, "extends": "llama", "is_modular": true, "parents": [ "llama", "gemma" ], "root": "llama" }, "blt": { "children": [], "depth": 1, "diff_size": 57, "extends": "mllama", "is_modular": true, "parents": [ "mllama", "llama" ], "root": "mllama" }, "camembert": { "children": [], "depth": 2, "diff_size": 15, "extends": "roberta", "is_modular": true, "parents": [ "roberta" ], "root": "bert" }, "chameleon": { "children": [ "janus" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "chameleon" }, "chinese_clip": { "children": [], "depth": 1, "diff_size": 46, "extends": "clip", "is_modular": true, "parents": [ "clip", "align", "bert" ], "root": "clip" }, "chmv2": { "children": [], "depth": 1, "diff_size": 85, "extends": "dpt", "is_modular": true, "parents": [ "depth_anything", "dpt" ], "root": "dpt" }, "clip": { "children": [ "altclip", "chinese_clip", "clipseg", "metaclip_2", "mlcd", "pp_ocrv5_server_rec", "x_clip" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "clip" }, "clipseg": { "children": [], "depth": 1, "diff_size": 56, "extends": "clip", "is_modular": true, "parents": [ "clip" ], "root": "clip" }, "cohere": { "children": [ "cohere2" ], "depth": 1, "diff_size": 17, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "cohere2": { "children": [ "cohere2_moe" ], "depth": 2, "diff_size": 38, "extends": "cohere", "is_modular": true, "parents": [ "cohere", "gemma2" ], "root": "llama" }, "cohere2_moe": { "children": [], "depth": 3, "diff_size": 20, "extends": "cohere2", "is_modular": true, "parents": [ "cohere2", "llama", "mixtral" ], "root": "llama" }, "cohere2_vision": { "children": [], "depth": 2, "diff_size": 26, "extends": "aya_vision", "is_modular": true, "parents": [ "aya_vision" ], "root": "llava" }, "cohere_asr": { "children": [], "depth": 3, "diff_size": 25, "extends": "moonshine", "is_modular": true, "parents": [ "moonshine", "clip" ], "root": "llama" }, "colmodernvbert": { "children": [], "depth": 1, "diff_size": 31, "extends": "colpali", "is_modular": true, "parents": [ "colpali" ], "root": "colpali" }, "colpali": { "children": [ "colmodernvbert", "colqwen2" ], "depth": 0, "diff_size": 13, "extends": null, "is_modular": true, "parents": [], "root": "colpali" }, "colqwen2": { "children": [], "depth": 1, "diff_size": 21, "extends": "colpali", "is_modular": true, "parents": [ "colpali" ], "root": "colpali" }, "conditional_detr": { "children": [], "depth": 1, "diff_size": 36, "extends": "detr", "is_modular": true, "parents": [ "detr" ], "root": "detr" }, "cosmos3_omni": { "children": [], "depth": 2, "diff_size": 10, "extends": "qwen3_vl", "is_modular": true, "parents": [ "qwen3_vl" ], "root": "qwen2_vl" }, "csm": { "children": [], "depth": 1, "diff_size": 62, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "cwm": { "children": [], "depth": 1, "diff_size": 37, "extends": "llama", "is_modular": true, "parents": [ "llama", "qwen2" ], "root": "llama" }, "d_fine": { "children": [ "deimv2" ], "depth": 2, "diff_size": 128, "extends": "rt_detr", "is_modular": true, "parents": [ "rt_detr" ], "root": "detr" }, "data2vec": { "children": [], "depth": 1, "diff_size": 35, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "dbrx": { "children": [], "depth": 1, "diff_size": 72, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "deepseek_ocr2": { "children": [], "depth": 2, "diff_size": 116, "extends": "qwen2", "is_modular": true, "parents": [ "qwen2", "deepseek_v2", "llava_next", "sam", "llama" ], "root": "llama" }, "deepseek_v2": { "children": [], "depth": 1, "diff_size": 59, "extends": "llama", "is_modular": true, "parents": [ "llama", "qwen2_moe" ], "root": "llama" }, "deepseek_v3": { "children": [ "deepseek_v32", "glm4_moe", "longcat_flash", "nemotron_h" ], "depth": 1, "diff_size": 18, "extends": "llama", "is_modular": true, "parents": [ "llama", "deepseek_v2", "qwen2_moe" ], "root": "llama" }, "deepseek_v32": { "children": [ "glm_moe_dsa" ], "depth": 2, "diff_size": 60, "extends": "deepseek_v3", "is_modular": true, "parents": [ "deepseek_v3", "glm4_moe_lite" ], "root": "llama" }, "deepseek_v4": { "children": [], "depth": 3, "diff_size": 98, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama", "deepseek_v3", "laguna" ], "root": "llama" }, "deepseek_vl": { "children": [ "deepseek_vl_hybrid" ], "depth": 2, "diff_size": 43, "extends": "janus", "is_modular": true, "parents": [ "janus", "idefics" ], "root": "chameleon" }, "deepseek_vl_hybrid": { "children": [], "depth": 3, "diff_size": 69, "extends": "deepseek_vl", "is_modular": true, "parents": [ "deepseek_vl", "idefics", "sam" ], "root": "chameleon" }, "deformable_detr": { "children": [ "lw_detr" ], "depth": 1, "diff_size": 91, "extends": "detr", "is_modular": true, "parents": [ "detr" ], "root": "detr" }, "deimv2": { "children": [], "depth": 3, "diff_size": 67, "extends": "d_fine", "is_modular": true, "parents": [ "d_fine", "llama" ], "root": "detr" }, "deit": { "children": [], "depth": 1, "diff_size": 19, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "depth_anything": { "children": [ "prompt_depth_anything" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "depth_anything" }, "detr": { "children": [ "conditional_detr", "deformable_detr", "maskformer", "rt_detr" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "detr" }, "dia": { "children": [], "depth": 1, "diff_size": 58, "extends": "llama", "is_modular": true, "parents": [ "llama", "phi3" ], "root": "llama" }, "diffllama": { "children": [], "depth": 1, "diff_size": 18, "extends": "llama", "is_modular": true, "parents": [ "llama", "gemma", "mistral" ], "root": "llama" }, "diffusion_gemma": { "children": [], "depth": 5, "diff_size": 106, "extends": "gemma4", "is_modular": true, "parents": [ "gemma4", "t5gemma2" ], "root": "llama" }, "dinov2": { "children": [ "dinov2_with_registers", "dinov3_vit", "eomt" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "dinov2" }, "dinov2_with_registers": { "children": [], "depth": 1, "diff_size": 37, "extends": "dinov2", "is_modular": true, "parents": [ "dinov2" ], "root": "dinov2" }, "dinov3_vit": { "children": [ "eomt_dinov3", "sapiens2" ], "depth": 1, "diff_size": 42, "extends": "dinov2", "is_modular": true, "parents": [ "dinov2", "arcee", "llama", "pixtral", "swin" ], "root": "dinov2" }, "doge": { "children": [], "depth": 1, "diff_size": 58, "extends": "llama", "is_modular": true, "parents": [ "llama", "mixtral" ], "root": "llama" }, "dots1": { "children": [], "depth": 3, "diff_size": 42, "extends": "qwen3", "is_modular": true, "parents": [ "deepseek_v3", "qwen3" ], "root": "llama" }, "dpt": { "children": [ "chmv2" ], "depth": 0, "diff_size": 29, "extends": null, "is_modular": true, "parents": [], "root": "dpt" }, "edgetam": { "children": [], "depth": 2, "diff_size": 37, "extends": "sam2", "is_modular": true, "parents": [ "sam2" ], "root": "sam" }, "edgetam_video": { "children": [], "depth": 3, "diff_size": 120, "extends": "sam2_video", "is_modular": true, "parents": [ "sam2_video" ], "root": "sam" }, "efficientloftr": { "children": [], "depth": 0, "diff_size": 8, "extends": null, "is_modular": true, "parents": [], "root": "efficientloftr" }, "emu3": { "children": [], "depth": 1, "diff_size": 138, "extends": "llama", "is_modular": true, "parents": [ "llama", "chameleon", "siglip" ], "root": "llama" }, "eomt": { "children": [ "videomt" ], "depth": 1, "diff_size": 89, "extends": "dinov2", "is_modular": true, "parents": [ "dinov2", "mask2former", "siglip" ], "root": "dinov2" }, "eomt_dinov3": { "children": [], "depth": 2, "diff_size": 53, "extends": "dinov3_vit", "is_modular": true, "parents": [ "dinov3_vit", "eomt" ], "root": "dinov2" }, "ernie": { "children": [], "depth": 1, "diff_size": 27, "extends": "bert", "is_modular": true, "parents": [ "bert" ], "root": "bert" }, "ernie4_5": { "children": [], "depth": 1, "diff_size": 4, "extends": "llama", "is_modular": true, "parents": [ "llama", "olmo" ], "root": "llama" }, "ernie4_5_moe": { "children": [ "ernie4_5_vl_moe" ], "depth": 3, "diff_size": 33, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama", "qwen3_moe", "ernie4_5" ], "root": "llama" }, "ernie4_5_vl_moe": { "children": [], "depth": 4, "diff_size": 168, "extends": "ernie4_5_moe", "is_modular": true, "parents": [ "ernie4_5_moe", "qwen2_5_vl", "qwen2_vl", "glm4v" ], "root": "llama" }, "esm": { "children": [ "evolla" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "esm" }, "eurobert": { "children": [], "depth": 1, "diff_size": 50, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "evolla": { "children": [], "depth": 1, "diff_size": 75, "extends": "esm", "is_modular": true, "parents": [ "esm", "llama" ], "root": "esm" }, "exaone4": { "children": [ "exaone_moe" ], "depth": 1, "diff_size": 38, "extends": "llama", "is_modular": true, "parents": [ "llama", "olmo2", "gemma2" ], "root": "llama" }, "exaone4_5": { "children": [], "depth": 2, "diff_size": 44, "extends": "qwen2_5_vl", "is_modular": true, "parents": [ "qwen2_5_vl", "exaone4" ], "root": "qwen2_vl" }, "exaone_moe": { "children": [], "depth": 2, "diff_size": 45, "extends": "exaone4", "is_modular": true, "parents": [ "exaone4", "deepseek_v3", "olmoe", "qwen2_moe" ], "root": "llama" }, "falcon_h1": { "children": [], "depth": 1, "diff_size": 39, "extends": "llama", "is_modular": true, "parents": [ "llama", "mamba2" ], "root": "llama" }, "falcon_mamba": { "children": [], "depth": 1, "diff_size": 16, "extends": "mamba", "is_modular": true, "parents": [ "mamba" ], "root": "mamba" }, "fast_vlm": { "children": [], "depth": 1, "diff_size": 19, "extends": "llava", "is_modular": true, "parents": [ "llava" ], "root": "llava" }, "fastspeech2_conformer": { "children": [ "parakeet" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "fastspeech2_conformer" }, "flex_olmo": { "children": [], "depth": 2, "diff_size": 38, "extends": "olmoe", "is_modular": true, "parents": [ "olmoe", "olmo2", "mixtral" ], "root": "llama" }, "florence2": { "children": [], "depth": 1, "diff_size": 145, "extends": "llava", "is_modular": true, "parents": [ "llava", "llama4", "swin" ], "root": "llava" }, "gemma": { "children": [ "gemma2" ], "depth": 1, "diff_size": 45, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "gemma2": { "children": [ "gemma3", "t5gemma", "vaultgemma" ], "depth": 2, "diff_size": 48, "extends": "gemma", "is_modular": true, "parents": [ "gemma" ], "root": "llama" }, "gemma3": { "children": [ "gemma3n", "gemma4", "modernbert", "t5gemma2" ], "depth": 3, "diff_size": 79, "extends": "gemma2", "is_modular": true, "parents": [ "gemma2", "paligemma" ], "root": "llama" }, "gemma3n": { "children": [], "depth": 4, "diff_size": 211, "extends": "gemma3", "is_modular": true, "parents": [ "gemma3", "paligemma", "gemma2" ], "root": "llama" }, "gemma4": { "children": [ "diffusion_gemma", "gemma4_unified" ], "depth": 4, "diff_size": 144, "extends": "gemma3", "is_modular": true, "parents": [ "gemma3", "gemma3n", "llama", "mixtral" ], "root": "llama" }, "gemma4_assistant": { "children": [ "gemma4_unified_assistant" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "gemma4_assistant" }, "gemma4_unified": { "children": [], "depth": 5, "diff_size": 107, "extends": "gemma4", "is_modular": true, "parents": [ "gemma4", "gemma2", "llama" ], "root": "llama" }, "gemma4_unified_assistant": { "children": [], "depth": 1, "diff_size": 1, "extends": "gemma4_assistant", "is_modular": true, "parents": [ "gemma4_assistant" ], "root": "gemma4_assistant" }, "glm": { "children": [ "glm4", "moonshine" ], "depth": 1, "diff_size": 2, "extends": "llama", "is_modular": true, "parents": [ "llama", "phi3" ], "root": "llama" }, "glm4": { "children": [], "depth": 2, "diff_size": 6, "extends": "glm", "is_modular": true, "parents": [ "glm", "phi3" ], "root": "llama" }, "glm46v": { "children": [], "depth": 3, "diff_size": 35, "extends": "glm4v", "is_modular": true, "parents": [ "glm4v" ], "root": "qwen2_vl" }, "glm4_moe": { "children": [ "glm4_moe_lite", "glm4v_moe", "solar_open" ], "depth": 2, "diff_size": 41, "extends": "deepseek_v3", "is_modular": true, "parents": [ "deepseek_v3", "cohere", "glm" ], "root": "llama" }, "glm4_moe_lite": { "children": [], "depth": 3, "diff_size": 46, "extends": "glm4_moe", "is_modular": true, "parents": [ "glm4_moe", "deepseek_v3" ], "root": "llama" }, "glm4v": { "children": [ "glm46v", "glm_image", "glm_ocr" ], "depth": 2, "diff_size": 122, "extends": "qwen2_5_vl", "is_modular": true, "parents": [ "qwen2_5_vl", "glm4", "qwen2_vl" ], "root": "qwen2_vl" }, "glm4v_moe": { "children": [], "depth": 3, "diff_size": 36, "extends": "glm4_moe", "is_modular": true, "parents": [ "glm4_moe", "glm4v", "qwen3_vl_moe", "deepseek_v3", "glm4" ], "root": "llama" }, "glm_image": { "children": [], "depth": 3, "diff_size": 120, "extends": "glm4v", "is_modular": true, "parents": [ "glm4v", "chameleon", "glm4v_moe", "siglip" ], "root": "qwen2_vl" }, "glm_moe_dsa": { "children": [], "depth": 3, "diff_size": 33, "extends": "deepseek_v32", "is_modular": true, "parents": [ "deepseek_v32", "deepseek_v3" ], "root": "llama" }, "glm_ocr": { "children": [], "depth": 3, "diff_size": 36, "extends": "glm4v", "is_modular": true, "parents": [ "glm4v" ], "root": "qwen2_vl" }, "glmasr": { "children": [], "depth": 3, "diff_size": 36, "extends": "audioflamingo3", "is_modular": true, "parents": [ "audioflamingo3", "glm", "llama" ], "root": "qwen2_audio" }, "glmga": { "children": [], "depth": 0, "diff_size": 40, "extends": null, "is_modular": true, "parents": [], "root": "glmga" }, "got_ocr2": { "children": [ "slanext" ], "depth": 1, "diff_size": 49, "extends": "llava", "is_modular": true, "parents": [ "llava", "sam" ], "root": "llava" }, "gpt_neox": { "children": [], "depth": 1, "diff_size": 49, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "gpt_oss": { "children": [ "openai_privacy_filter" ], "depth": 3, "diff_size": 28, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama", "qwen2" ], "root": "llama" }, "granite": { "children": [ "hyperclovax" ], "depth": 1, "diff_size": 7, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "granite4_vision": { "children": [], "depth": 1, "diff_size": 49, "extends": "llava_next", "is_modular": true, "parents": [ "llava_next", "granite" ], "root": "llava_next" }, "granite_speech": { "children": [ "granite_speech_plus" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "granite_speech" }, "granite_speech_plus": { "children": [], "depth": 1, "diff_size": 9, "extends": "granite_speech", "is_modular": true, "parents": [ "granite_speech" ], "root": "granite_speech" }, "granitemoe": { "children": [ "granitemoeshared" ], "depth": 3, "diff_size": 21, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "granite", "jetmoe", "llama" ], "root": "llama" }, "granitemoehybrid": { "children": [], "depth": 5, "diff_size": 17, "extends": "granitemoeshared", "is_modular": true, "parents": [ "granitemoeshared", "bamba", "gemma2" ], "root": "llama" }, "granitemoeshared": { "children": [ "granitemoehybrid" ], "depth": 4, "diff_size": 20, "extends": "granitemoe", "is_modular": true, "parents": [ "granitemoe" ], "root": "llama" }, "grounding_dino": { "children": [ "mm_grounding_dino" ], "depth": 0, "diff_size": 14, "extends": null, "is_modular": true, "parents": [], "root": "grounding_dino" }, "helium": { "children": [], "depth": 1, "diff_size": 9, "extends": "llama", "is_modular": true, "parents": [ "llama", "gemma", "granite" ], "root": "llama" }, "hgnet_v2": { "children": [], "depth": 2, "diff_size": 78, "extends": "rt_detr", "is_modular": true, "parents": [ "rt_detr" ], "root": "detr" }, "higgs_audio_v2": { "children": [], "depth": 1, "diff_size": 36, "extends": "llama", "is_modular": true, "parents": [ "llama", "csm" ], "root": "llama" }, "higgs_audio_v2_tokenizer": { "children": [], "depth": 1, "diff_size": 19, "extends": "xcodec", "is_modular": true, "parents": [ "xcodec" ], "root": "xcodec" }, "hrm_text": { "children": [], "depth": 1, "diff_size": 38, "extends": "llama", "is_modular": true, "parents": [ "llama", "nanochat" ], "root": "llama" }, "hubert": { "children": [], "depth": 1, "diff_size": 28, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "hunyuan_v1_dense": { "children": [], "depth": 1, "diff_size": 6, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "hunyuan_v1_moe": { "children": [], "depth": 1, "diff_size": 15, "extends": "llama", "is_modular": true, "parents": [ "llama", "hunyuan_v1_dense", "mixtral" ], "root": "llama" }, "hy_v3": { "children": [], "depth": 1, "diff_size": 52, "extends": "llama", "is_modular": true, "parents": [ "llama", "minimax_m2", "apertus", "deepseek_v3", "mixtral", "qwen3_moe" ], "root": "llama" }, "hyperclovax": { "children": [], "depth": 2, "diff_size": 12, "extends": "granite", "is_modular": true, "parents": [ "granite" ], "root": "llama" }, "idefics3": { "children": [ "smolvlm" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "idefics3" }, "ijepa": { "children": [], "depth": 1, "diff_size": 7, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "informer": { "children": [], "depth": 1, "diff_size": 31, "extends": "time_series_transformer", "is_modular": true, "parents": [ "time_series_transformer", "bart" ], "root": "time_series_transformer" }, "instructblip": { "children": [ "instructblipvideo" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "instructblip" }, "instructblipvideo": { "children": [], "depth": 1, "diff_size": 20, "extends": "instructblip", "is_modular": true, "parents": [ "instructblip" ], "root": "instructblip" }, "internvl": { "children": [ "qianfan_ocr" ], "depth": 1, "diff_size": 57, "extends": "llava", "is_modular": true, "parents": [ "llava", "clip", "janus", "llama" ], "root": "llava" }, "jais2": { "children": [], "depth": 1, "diff_size": 20, "extends": "llama", "is_modular": true, "parents": [ "llama", "nemotron" ], "root": "llama" }, "jamba": { "children": [], "depth": 1, "diff_size": 49, "extends": "llama", "is_modular": true, "parents": [ "llama", "mixtral", "mistral" ], "root": "llama" }, "janus": { "children": [ "deepseek_vl" ], "depth": 1, "diff_size": 147, "extends": "chameleon", "is_modular": true, "parents": [ "chameleon", "siglip", "idefics", "blip_2" ], "root": "chameleon" }, "jetmoe": { "children": [], "depth": 3, "diff_size": 50, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama" ], "root": "llama" }, "jina_embeddings_v3": { "children": [], "depth": 3, "diff_size": 30, "extends": "xlm_roberta", "is_modular": true, "parents": [ "xlm_roberta", "llama", "clip", "gpt_neox" ], "root": "bert" }, "kyutai_speech_to_text": { "children": [], "depth": 1, "diff_size": 23, "extends": "moshi", "is_modular": true, "parents": [ "moshi", "llama", "mimi" ], "root": "moshi" }, "laguna": { "children": [], "depth": 4, "diff_size": 45, "extends": "qwen2_moe", "is_modular": true, "parents": [ "qwen2_moe", "qwen3_moe", "afmoe", "gemma3", "glm4_moe_lite", "llama", "qwen3_5_moe" ], "root": "llama" }, "lasr": { "children": [], "depth": 2, "diff_size": 61, "extends": "parakeet", "is_modular": true, "parents": [ "parakeet", "llama" ], "root": "fastspeech2_conformer" }, "layoutlmv3": { "children": [ "pp_doclayout_v2" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "layoutlmv3" }, "layoutxlm": { "children": [], "depth": 0, "diff_size": 3, "extends": null, "is_modular": true, "parents": [], "root": "layoutxlm" }, "lfm2": { "children": [ "lfm2_moe" ], "depth": 1, "diff_size": 23, "extends": "llama", "is_modular": true, "parents": [ "llama", "gemma2" ], "root": "llama" }, "lfm2_moe": { "children": [], "depth": 2, "diff_size": 12, "extends": "lfm2", "is_modular": true, "parents": [ "lfm2", "llama", "qwen2_moe", "mixtral", "qwen3_moe" ], "root": "llama" }, "lfm2_vl": { "children": [], "depth": 1, "diff_size": 14, "extends": "llava", "is_modular": true, "parents": [ "llava" ], "root": "llava" }, "lightglue": { "children": [], "depth": 1, "diff_size": 90, "extends": "llama", "is_modular": true, "parents": [ "clip", "llama" ], "root": "llama" }, "lighton_ocr": { "children": [], "depth": 2, "diff_size": 28, "extends": "mistral3", "is_modular": true, "parents": [ "mistral3" ], "root": "llava" }, "llama": { "children": [ "afmoe", "apertus", "arcee", "aria", "bamba", "bitnet", "cohere", "csm", "cwm", "dbrx", "deepseek_v2", "deepseek_v3", "dia", "diffllama", "doge", "emu3", "ernie4_5", "eurobert", "exaone4", "falcon_h1", "gemma", "glm", "gpt_neox", "granite", "helium", "higgs_audio_v2", "hrm_text", "hunyuan_v1_dense", "hunyuan_v1_moe", "hy_v3", "jais2", "jamba", "lfm2", "lightglue", "minicpm3", "mistral", "mistral4", "nanochat", "olmo", "olmoe", "phi", "qwen2", "seed_oss", "smollm3", "timesfm", "vibevoice_acoustic_tokenizer", "youtu" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "llama" }, "llava": { "children": [ "aya_vision", "fast_vlm", "florence2", "got_ocr2", "internvl", "lfm2_vl", "mistral3", "perception_lm", "vipllava" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "llava" }, "llava_next": { "children": [ "granite4_vision", "llava_next_video", "ovis2" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "llava_next" }, "llava_next_video": { "children": [ "llava_onevision" ], "depth": 1, "diff_size": 38, "extends": "llava_next", "is_modular": true, "parents": [ "llava_next" ], "root": "llava_next" }, "llava_onevision": { "children": [], "depth": 2, "diff_size": 52, "extends": "llava_next_video", "is_modular": true, "parents": [ "llava_next_video" ], "root": "llava_next" }, "longcat_flash": { "children": [], "depth": 2, "diff_size": 41, "extends": "deepseek_v3", "is_modular": true, "parents": [ "deepseek_v3", "mixtral" ], "root": "llama" }, "lw_detr": { "children": [ "rf_detr" ], "depth": 2, "diff_size": 177, "extends": "deformable_detr", "is_modular": true, "parents": [ "deformable_detr", "vitdet", "convnext", "rt_detr", "vit" ], "root": "detr" }, "mamba": { "children": [ "falcon_mamba" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "mamba" }, "mask2former": { "children": [], "depth": 0, "diff_size": 12, "extends": null, "is_modular": true, "parents": [], "root": "mask2former" }, "maskformer": { "children": [], "depth": 1, "diff_size": 143, "extends": "detr", "is_modular": true, "parents": [ "detr" ], "root": "detr" }, "mellum": { "children": [], "depth": 5, "diff_size": 25, "extends": "qwen3_moe", "is_modular": true, "parents": [ "qwen3_moe", "laguna" ], "root": "llama" }, "metaclip_2": { "children": [], "depth": 1, "diff_size": 19, "extends": "clip", "is_modular": true, "parents": [ "clip" ], "root": "clip" }, "minicpm3": { "children": [], "depth": 1, "diff_size": 31, "extends": "llama", "is_modular": true, "parents": [ "llama", "deepseek_v2", "gemma3" ], "root": "llama" }, "minicpmv4_6": { "children": [], "depth": 1, "diff_size": 88, "extends": "siglip", "is_modular": true, "parents": [ "siglip", "idefics3", "lfm2_vl", "qwen2_vl" ], "root": "siglip" }, "minimax": { "children": [], "depth": 3, "diff_size": 67, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "gemma2" ], "root": "llama" }, "minimax_m2": { "children": [ "minimax_m3_vl" ], "depth": 3, "diff_size": 39, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "flex_olmo", "glm4_moe" ], "root": "llama" }, "minimax_m3_vl": { "children": [], "depth": 4, "diff_size": 173, "extends": "minimax_m2", "is_modular": true, "parents": [ "minimax_m2", "llava", "clip", "deepseek_v4", "gemma3", "laguna", "mixtral", "qwen2_5_vl" ], "root": "llama" }, "ministral": { "children": [], "depth": 2, "diff_size": 9, "extends": "qwen2", "is_modular": true, "parents": [ "qwen2" ], "root": "llama" }, "ministral3": { "children": [], "depth": 2, "diff_size": 10, "extends": "mistral", "is_modular": true, "parents": [ "mistral" ], "root": "llama" }, "mistral": { "children": [ "ministral3", "mixtral", "phi3", "starcoder2", "voxtral_realtime" ], "depth": 1, "diff_size": 9, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "mistral3": { "children": [ "lighton_ocr" ], "depth": 1, "diff_size": 14, "extends": "llava", "is_modular": true, "parents": [ "llava", "mistral" ], "root": "llava" }, "mistral4": { "children": [], "depth": 1, "diff_size": 28, "extends": "llama", "is_modular": true, "parents": [ "deepseek_v3", "llama", "deepseek_v2", "qwen2_moe" ], "root": "llama" }, "mixtral": { "children": [ "deepseek_v4", "ernie4_5_moe", "gpt_oss", "granitemoe", "jetmoe", "minimax", "minimax_m2", "phimoe", "qwen2_moe" ], "depth": 2, "diff_size": 26, "extends": "mistral", "is_modular": true, "parents": [ "mistral" ], "root": "llama" }, "mlcd": { "children": [], "depth": 1, "diff_size": 43, "extends": "clip", "is_modular": true, "parents": [ "clip", "qwen2_vl" ], "root": "clip" }, "mllama": { "children": [ "blt" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "mllama" }, "mm_grounding_dino": { "children": [], "depth": 1, "diff_size": 51, "extends": "grounding_dino", "is_modular": true, "parents": [ "grounding_dino" ], "root": "grounding_dino" }, "modernbert": { "children": [ "modernbert_decoder" ], "depth": 4, "diff_size": 118, "extends": "gemma3", "is_modular": true, "parents": [ "gemma3" ], "root": "llama" }, "modernbert_decoder": { "children": [], "depth": 5, "diff_size": 74, "extends": "modernbert", "is_modular": true, "parents": [ "modernbert" ], "root": "llama" }, "modernvbert": { "children": [], "depth": 2, "diff_size": 60, "extends": "smolvlm", "is_modular": true, "parents": [ "smolvlm", "modernbert" ], "root": "idefics3" }, "moonshine": { "children": [ "cohere_asr", "moonshine_streaming" ], "depth": 2, "diff_size": 90, "extends": "glm", "is_modular": true, "parents": [ "glm", "llama", "whisper" ], "root": "llama" }, "moonshine_streaming": { "children": [], "depth": 3, "diff_size": 54, "extends": "moonshine", "is_modular": true, "parents": [ "moonshine", "llama" ], "root": "llama" }, "moshi": { "children": [ "kyutai_speech_to_text" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "moshi" }, "musicflamingo": { "children": [], "depth": 3, "diff_size": 29, "extends": "audioflamingo3", "is_modular": true, "parents": [ "audioflamingo3", "moonshine" ], "root": "qwen2_audio" }, "nanochat": { "children": [], "depth": 1, "diff_size": 9, "extends": "llama", "is_modular": true, "parents": [ "llama", "clip", "gemma2", "llama4", "qwen3" ], "root": "llama" }, "nemotron3_5_asr": { "children": [], "depth": 3, "diff_size": 29, "extends": "nemotron_asr_streaming", "is_modular": true, "parents": [ "nemotron_asr_streaming" ], "root": "fastspeech2_conformer" }, "nemotron_asr_streaming": { "children": [ "nemotron3_5_asr" ], "depth": 2, "diff_size": 73, "extends": "parakeet", "is_modular": true, "parents": [ "parakeet", "voxtral_realtime", "fastspeech2_conformer" ], "root": "fastspeech2_conformer" }, "nemotron_h": { "children": [], "depth": 2, "diff_size": 43, "extends": "deepseek_v3", "is_modular": true, "parents": [ "deepseek_v3", "jamba", "llama", "nemotron", "zamba", "zamba2" ], "root": "llama" }, "nomic_bert": { "children": [], "depth": 1, "diff_size": 40, "extends": "bert", "is_modular": true, "parents": [ "bert", "jina_embeddings_v3", "gemma", "llama" ], "root": "bert" }, "olmo": { "children": [ "olmo2" ], "depth": 1, "diff_size": 10, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "olmo2": { "children": [ "olmo3" ], "depth": 2, "diff_size": 14, "extends": "olmo", "is_modular": true, "parents": [ "olmo", "llama" ], "root": "llama" }, "olmo3": { "children": [ "olmo_hybrid" ], "depth": 3, "diff_size": 20, "extends": "olmo2", "is_modular": true, "parents": [ "olmo2", "gemma3" ], "root": "llama" }, "olmo_hybrid": { "children": [], "depth": 4, "diff_size": 66, "extends": "olmo3", "is_modular": true, "parents": [ "olmo3", "qwen3_next", "llama", "olmo2" ], "root": "llama" }, "olmoe": { "children": [ "flex_olmo" ], "depth": 1, "diff_size": 27, "extends": "llama", "is_modular": true, "parents": [ "llama", "mixtral", "gemma", "qwen2_moe" ], "root": "llama" }, "openai_privacy_filter": { "children": [], "depth": 4, "diff_size": 41, "extends": "gpt_oss", "is_modular": true, "parents": [ "gpt_oss" ], "root": "llama" }, "ovis2": { "children": [], "depth": 1, "diff_size": 48, "extends": "llava_next", "is_modular": true, "parents": [ "aimv2", "llama", "llava", "llava_next", "siglip" ], "root": "llava_next" }, "owlv2": { "children": [], "depth": 0, "diff_size": 35, "extends": null, "is_modular": true, "parents": [], "root": "owlv2" }, "paddleocr_vl": { "children": [], "depth": 1, "diff_size": 112, "extends": "qwen2_vl", "is_modular": true, "parents": [ "qwen2_vl", "ernie4_5", "video_llama_3", "siglip", "qwen2_5_omni" ], "root": "qwen2_vl" }, "parakeet": { "children": [ "lasr", "nemotron_asr_streaming" ], "depth": 1, "diff_size": 110, "extends": "fastspeech2_conformer", "is_modular": true, "parents": [ "fastspeech2_conformer", "llama" ], "root": "fastspeech2_conformer" }, "pe_audio": { "children": [], "depth": 4, "diff_size": 39, "extends": "pe_audio_video", "is_modular": true, "parents": [ "pe_audio_video", "dac" ], "root": "llama" }, "pe_audio_video": { "children": [ "pe_audio", "pe_video" ], "depth": 3, "diff_size": 111, "extends": "qwen3", "is_modular": true, "parents": [ "qwen3" ], "root": "llama" }, "pe_video": { "children": [], "depth": 4, "diff_size": 28, "extends": "pe_audio_video", "is_modular": true, "parents": [ "pe_audio_video" ], "root": "llama" }, "perception_lm": { "children": [], "depth": 1, "diff_size": 20, "extends": "llava", "is_modular": true, "parents": [ "llava" ], "root": "llava" }, "phi": { "children": [], "depth": 1, "diff_size": 12, "extends": "llama", "is_modular": true, "parents": [ "llama", "clip" ], "root": "llama" }, "phi3": { "children": [], "depth": 2, "diff_size": 14, "extends": "mistral", "is_modular": true, "parents": [ "mistral", "phi" ], "root": "llama" }, "phi4_multimodal": { "children": [], "depth": 1, "diff_size": 174, "extends": "siglip", "is_modular": true, "parents": [ "siglip", "phi3" ], "root": "siglip" }, "phimoe": { "children": [], "depth": 3, "diff_size": 25, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama" ], "root": "llama" }, "pi0": { "children": [], "depth": 0, "diff_size": 77, "extends": null, "is_modular": true, "parents": [], "root": "pi0" }, "pixio": { "children": [], "depth": 1, "diff_size": 31, "extends": "vit", "is_modular": true, "parents": [ "vit", "dinov2", "swin" ], "root": "vit" }, "plbart": { "children": [], "depth": 1, "diff_size": 32, "extends": "bart", "is_modular": true, "parents": [ "bart", "bigbird_pegasus" ], "root": "bart" }, "pp_chart2table": { "children": [], "depth": 0, "diff_size": 29, "extends": null, "is_modular": true, "parents": [], "root": "pp_chart2table" }, "pp_doclayout_v2": { "children": [], "depth": 1, "diff_size": 149, "extends": "layoutlmv3", "is_modular": true, "parents": [ "layoutlmv3", "rt_detr", "pp_doclayout_v3" ], "root": "layoutlmv3" }, "pp_doclayout_v3": { "children": [], "depth": 2, "diff_size": 133, "extends": "rt_detr", "is_modular": true, "parents": [ "rt_detr", "resnet" ], "root": "detr" }, "pp_formulanet": { "children": [], "depth": 3, "diff_size": 73, "extends": "slanext", "is_modular": true, "parents": [ "slanext", "florence2", "mbart" ], "root": "llava" }, "pp_lcnet": { "children": [ "pp_lcnet_v3", "pp_lcnet_v4", "uvdoc" ], "depth": 1, "diff_size": 83, "extends": "resnet", "is_modular": true, "parents": [ "resnet" ], "root": "resnet" }, "pp_lcnet_v3": { "children": [], "depth": 2, "diff_size": 21, "extends": "pp_lcnet", "is_modular": true, "parents": [ "pp_lcnet", "hgnet_v2" ], "root": "resnet" }, "pp_lcnet_v4": { "children": [], "depth": 2, "diff_size": 32, "extends": "pp_lcnet", "is_modular": true, "parents": [ "pp_lcnet", "hgnet_v2" ], "root": "resnet" }, "pp_ocrv5_mobile_det": { "children": [ "pp_ocrv6_small_det" ], "depth": 1, "diff_size": 33, "extends": "pp_ocrv5_server_det", "is_modular": true, "parents": [ "pp_ocrv5_server_det" ], "root": "pp_ocrv5_server_det" }, "pp_ocrv5_mobile_rec": { "children": [], "depth": 2, "diff_size": 5, "extends": "pp_ocrv5_server_rec", "is_modular": true, "parents": [ "pp_ocrv5_server_rec" ], "root": "clip" }, "pp_ocrv5_server_det": { "children": [ "pp_ocrv5_mobile_det", "pp_ocrv6_medium_det" ], "depth": 0, "diff_size": 93, "extends": null, "is_modular": true, "parents": [], "root": "pp_ocrv5_server_det" }, "pp_ocrv5_server_rec": { "children": [ "pp_ocrv5_mobile_rec", "pp_ocrv6_small_rec" ], "depth": 1, "diff_size": 75, "extends": "clip", "is_modular": true, "parents": [ "blip_2", "clip", "focalnet", "paddleocr_vl", "resnet" ], "root": "clip" }, "pp_ocrv6_medium_det": { "children": [], "depth": 1, "diff_size": 6, "extends": "pp_ocrv5_server_det", "is_modular": true, "parents": [ "pp_ocrv5_server_det", "pp_ocrv5_mobile_det" ], "root": "pp_ocrv5_server_det" }, "pp_ocrv6_small_det": { "children": [], "depth": 2, "diff_size": 8, "extends": "pp_ocrv5_mobile_det", "is_modular": true, "parents": [ "pp_ocrv5_mobile_det", "pp_lcnet", "pp_ocrv5_server_det" ], "root": "pp_ocrv5_server_det" }, "pp_ocrv6_small_rec": { "children": [ "pp_ocrv6_tiny_rec" ], "depth": 2, "diff_size": 12, "extends": "pp_ocrv5_server_rec", "is_modular": true, "parents": [ "pp_ocrv5_server_rec" ], "root": "clip" }, "pp_ocrv6_tiny_rec": { "children": [], "depth": 3, "diff_size": 22, "extends": "pp_ocrv6_small_rec", "is_modular": true, "parents": [ "pp_ocrv6_small_rec" ], "root": "clip" }, "prompt_depth_anything": { "children": [], "depth": 1, "diff_size": 27, "extends": "depth_anything", "is_modular": true, "parents": [ "depth_anything" ], "root": "depth_anything" }, "qianfan_ocr": { "children": [], "depth": 2, "diff_size": 34, "extends": "internvl", "is_modular": true, "parents": [ "internvl", "beit" ], "root": "llava" }, "qwen2": { "children": [ "deepseek_ocr2", "ministral", "qwen3", "vibevoice_asr" ], "depth": 1, "diff_size": 5, "extends": "llama", "is_modular": true, "parents": [ "llama", "gemma2", "mistral" ], "root": "llama" }, "qwen2_5_omni": { "children": [ "xcodec2" ], "depth": 2, "diff_size": 440, "extends": "qwen2_5_vl", "is_modular": true, "parents": [ "qwen2_5_vl", "llama", "qwen2_audio", "qwen2_vl" ], "root": "qwen2_vl" }, "qwen2_5_vl": { "children": [ "exaone4_5", "glm4v", "qwen2_5_omni" ], "depth": 1, "diff_size": 69, "extends": "qwen2_vl", "is_modular": true, "parents": [ "qwen2_vl", "llama" ], "root": "qwen2_vl" }, "qwen2_audio": { "children": [ "voxtral" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "qwen2_audio" }, "qwen2_moe": { "children": [ "laguna", "qwen3_moe" ], "depth": 3, "diff_size": 30, "extends": "mixtral", "is_modular": true, "parents": [ "mixtral", "llama", "gemma", "gemma2" ], "root": "llama" }, "qwen2_vl": { "children": [ "paddleocr_vl", "qwen2_5_vl", "qwen3_vl", "video_llama_3" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "qwen2_vl" }, "qwen3": { "children": [ "dots1", "pe_audio_video" ], "depth": 2, "diff_size": 3, "extends": "qwen2", "is_modular": true, "parents": [ "qwen2", "gemma", "llama" ], "root": "llama" }, "qwen3_5": { "children": [ "qwen3_5_moe" ], "depth": 2, "diff_size": 78, "extends": "qwen3_vl", "is_modular": true, "parents": [ "qwen3_next", "qwen3_vl", "qwen3" ], "root": "qwen2_vl" }, "qwen3_5_moe": { "children": [], "depth": 3, "diff_size": 38, "extends": "qwen3_5", "is_modular": true, "parents": [ "qwen3_5", "qwen3_next", "qwen3_vl_moe" ], "root": "qwen2_vl" }, "qwen3_asr": { "children": [], "depth": 6, "diff_size": 43, "extends": "qwen3_omni_moe", "is_modular": true, "parents": [ "audioflamingo3", "qwen3_omni_moe", "qwen2_audio", "voxtral" ], "root": "llama" }, "qwen3_moe": { "children": [ "mellum", "qwen3_next", "qwen3_omni_moe" ], "depth": 4, "diff_size": 9, "extends": "qwen2_moe", "is_modular": true, "parents": [ "llama", "qwen2_moe", "mixtral", "qwen3" ], "root": "llama" }, "qwen3_next": { "children": [], "depth": 5, "diff_size": 38, "extends": "qwen3_moe", "is_modular": true, "parents": [ "llama", "qwen2_moe", "qwen3_moe", "gemma2", "gemma3", "mixtral" ], "root": "llama" }, "qwen3_omni_moe": { "children": [ "qwen3_asr" ], "depth": 5, "diff_size": 316, "extends": "qwen3_moe", "is_modular": true, "parents": [ "qwen3", "qwen3_moe", "qwen2_5_omni", "qwen3_vl_moe", "qwen2_moe", "mimi" ], "root": "llama" }, "qwen3_vl": { "children": [ "cosmos3_omni", "qwen3_5", "qwen3_vl_moe" ], "depth": 1, "diff_size": 120, "extends": "qwen2_vl", "is_modular": true, "parents": [ "qwen2_vl", "qwen2_5_vl", "qwen3", "llama" ], "root": "qwen2_vl" }, "qwen3_vl_moe": { "children": [], "depth": 2, "diff_size": 48, "extends": "qwen3_vl", "is_modular": true, "parents": [ "qwen3_vl", "qwen3_moe" ], "root": "qwen2_vl" }, "resnet": { "children": [ "pp_lcnet" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "resnet" }, "rf_detr": { "children": [], "depth": 3, "diff_size": 106, "extends": "lw_detr", "is_modular": true, "parents": [ "lw_detr", "dinov2", "clip", "convnext" ], "root": "detr" }, "roberta": { "children": [ "camembert", "xlm_roberta" ], "depth": 1, "diff_size": 63, "extends": "bert", "is_modular": true, "parents": [ "bert" ], "root": "bert" }, "rt_detr": { "children": [ "d_fine", "hgnet_v2", "pp_doclayout_v3", "rt_detr_v2" ], "depth": 1, "diff_size": 197, "extends": "detr", "is_modular": true, "parents": [ "detr", "deformable_detr" ], "root": "detr" }, "rt_detr_v2": { "children": [], "depth": 2, "diff_size": 74, "extends": "rt_detr", "is_modular": true, "parents": [ "rt_detr" ], "root": "detr" }, "sam": { "children": [ "sam2", "sam_hq" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "sam" }, "sam2": { "children": [ "edgetam", "sam2_video", "sam3_tracker" ], "depth": 1, "diff_size": 131, "extends": "sam", "is_modular": true, "parents": [ "sam", "detr" ], "root": "sam" }, "sam2_video": { "children": [ "edgetam_video", "sam3_tracker_video" ], "depth": 2, "diff_size": 172, "extends": "sam2", "is_modular": true, "parents": [ "sam2" ], "root": "sam" }, "sam3": { "children": [], "depth": 0, "diff_size": 10, "extends": null, "is_modular": true, "parents": [], "root": "sam3" }, "sam3_lite_text": { "children": [], "depth": 1, "diff_size": 98, "extends": "siglip", "is_modular": true, "parents": [ "siglip", "sam3" ], "root": "siglip" }, "sam3_tracker": { "children": [], "depth": 2, "diff_size": 20, "extends": "sam2", "is_modular": true, "parents": [ "sam2" ], "root": "sam" }, "sam3_tracker_video": { "children": [], "depth": 3, "diff_size": 63, "extends": "sam2_video", "is_modular": true, "parents": [ "sam2_video" ], "root": "sam" }, "sam_hq": { "children": [], "depth": 1, "diff_size": 30, "extends": "sam", "is_modular": true, "parents": [ "sam" ], "root": "sam" }, "sapiens2": { "children": [], "depth": 2, "diff_size": 141, "extends": "dinov3_vit", "is_modular": true, "parents": [ "dinov3_vit", "llama", "mask2former", "pp_ocrv5_server_det", "vitmatte", "vitpose" ], "root": "dinov2" }, "seed_oss": { "children": [], "depth": 1, "diff_size": 11, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "segformer": { "children": [], "depth": 1, "diff_size": 99, "extends": "vit", "is_modular": true, "parents": [ "vit", "swin" ], "root": "vit" }, "sew": { "children": [], "depth": 1, "diff_size": 35, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "siglip": { "children": [ "aimv2", "minicpmv4_6", "phi4_multimodal", "sam3_lite_text", "siglip2" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "siglip" }, "siglip2": { "children": [], "depth": 1, "diff_size": 30, "extends": "siglip", "is_modular": true, "parents": [ "siglip" ], "root": "siglip" }, "slanet": { "children": [], "depth": 3, "diff_size": 43, "extends": "slanext", "is_modular": true, "parents": [ "slanext", "pp_lcnet" ], "root": "llava" }, "slanext": { "children": [ "pp_formulanet", "slanet" ], "depth": 2, "diff_size": 75, "extends": "got_ocr2", "is_modular": true, "parents": [ "got_ocr2" ], "root": "llava" }, "smollm3": { "children": [], "depth": 1, "diff_size": 35, "extends": "llama", "is_modular": true, "parents": [ "llama", "qwen2" ], "root": "llama" }, "smolvlm": { "children": [ "modernvbert" ], "depth": 1, "diff_size": 20, "extends": "idefics3", "is_modular": true, "parents": [ "idefics3" ], "root": "idefics3" }, "solar_open": { "children": [], "depth": 3, "diff_size": 19, "extends": "glm4_moe", "is_modular": true, "parents": [ "glm4_moe", "llama" ], "root": "llama" }, "starcoder2": { "children": [], "depth": 2, "diff_size": 10, "extends": "mistral", "is_modular": true, "parents": [ "mistral" ], "root": "llama" }, "swin": { "children": [], "depth": 1, "diff_size": 109, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "switch_transformers": { "children": [], "depth": 1, "diff_size": 67, "extends": "t5", "is_modular": true, "parents": [ "t5" ], "root": "t5" }, "t5": { "children": [ "switch_transformers" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "t5" }, "t5gemma": { "children": [], "depth": 3, "diff_size": 104, "extends": "gemma2", "is_modular": true, "parents": [ "gemma2" ], "root": "llama" }, "t5gemma2": { "children": [], "depth": 4, "diff_size": 122, "extends": "gemma3", "is_modular": true, "parents": [ "gemma3", "t5gemma" ], "root": "llama" }, "time_series_transformer": { "children": [ "informer" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "time_series_transformer" }, "timesfm": { "children": [ "timesfm2_5" ], "depth": 1, "diff_size": 66, "extends": "llama", "is_modular": true, "parents": [ "llama" ], "root": "llama" }, "timesfm2_5": { "children": [], "depth": 2, "diff_size": 47, "extends": "timesfm", "is_modular": true, "parents": [ "timesfm", "llama", "apertus", "clip" ], "root": "llama" }, "unispeech": { "children": [], "depth": 1, "diff_size": 36, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "unispeech_sat": { "children": [], "depth": 1, "diff_size": 38, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "uvdoc": { "children": [], "depth": 2, "diff_size": 83, "extends": "pp_lcnet", "is_modular": true, "parents": [ "pp_lcnet", "pp_ocrv5_server_det" ], "root": "resnet" }, "vaultgemma": { "children": [], "depth": 3, "diff_size": 7, "extends": "gemma2", "is_modular": true, "parents": [ "gemma2" ], "root": "llama" }, "vibevoice_acoustic_tokenizer": { "children": [], "depth": 1, "diff_size": 83, "extends": "llama", "is_modular": true, "parents": [ "llama", "voxtral_realtime" ], "root": "llama" }, "vibevoice_asr": { "children": [], "depth": 2, "diff_size": 58, "extends": "qwen2", "is_modular": true, "parents": [ "qwen2", "vibevoice_acoustic_tokenizer" ], "root": "llama" }, "video_llama_3": { "children": [], "depth": 1, "diff_size": 134, "extends": "qwen2_vl", "is_modular": true, "parents": [ "qwen2_vl", "siglip" ], "root": "qwen2_vl" }, "videomt": { "children": [], "depth": 2, "diff_size": 23, "extends": "eomt", "is_modular": true, "parents": [ "eomt" ], "root": "dinov2" }, "videoprism": { "children": [], "depth": 2, "diff_size": 145, "extends": "vivit", "is_modular": true, "parents": [ "vivit" ], "root": "vit" }, "vipllava": { "children": [], "depth": 1, "diff_size": 9, "extends": "llava", "is_modular": true, "parents": [ "llava" ], "root": "llava" }, "vit": { "children": [ "audio_spectrogram_transformer", "beit", "deit", "ijepa", "pixio", "segformer", "swin", "vit_mae", "vit_msn", "vivit" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "vit" }, "vit_mae": { "children": [], "depth": 1, "diff_size": 46, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "vit_msn": { "children": [], "depth": 1, "diff_size": 10, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "vivit": { "children": [ "videoprism" ], "depth": 1, "diff_size": 20, "extends": "vit", "is_modular": true, "parents": [ "vit" ], "root": "vit" }, "voxtral": { "children": [ "audioflamingo3" ], "depth": 1, "diff_size": 31, "extends": "qwen2_audio", "is_modular": true, "parents": [ "qwen2_audio" ], "root": "qwen2_audio" }, "voxtral_realtime": { "children": [], "depth": 2, "diff_size": 94, "extends": "mistral", "is_modular": true, "parents": [ "mistral", "voxtral", "llama" ], "root": "llama" }, "wav2vec2": { "children": [ "data2vec", "hubert", "sew", "unispeech", "unispeech_sat", "wav2vec2_conformer", "wavlm" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "wav2vec2" }, "wav2vec2_bert": { "children": [], "depth": 2, "diff_size": 63, "extends": "wav2vec2_conformer", "is_modular": true, "parents": [ "wav2vec2_conformer", "wav2vec2" ], "root": "wav2vec2" }, "wav2vec2_conformer": { "children": [ "wav2vec2_bert" ], "depth": 1, "diff_size": 63, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "wavlm": { "children": [], "depth": 1, "diff_size": 46, "extends": "wav2vec2", "is_modular": true, "parents": [ "wav2vec2" ], "root": "wav2vec2" }, "x_clip": { "children": [], "depth": 1, "diff_size": 49, "extends": "clip", "is_modular": true, "parents": [ "clip", "altclip", "beit" ], "root": "clip" }, "xcodec": { "children": [ "higgs_audio_v2_tokenizer" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "xcodec" }, "xcodec2": { "children": [], "depth": 3, "diff_size": 102, "extends": "qwen2_5_omni", "is_modular": true, "parents": [ "qwen2_5_omni", "dac", "llama", "clip", "voxtral" ], "root": "qwen2_vl" }, "xlm_roberta": { "children": [ "jina_embeddings_v3" ], "depth": 2, "diff_size": 15, "extends": "roberta", "is_modular": true, "parents": [ "roberta" ], "root": "bert" }, "xlm_roberta_xl": { "children": [], "depth": 1, "diff_size": 63, "extends": "bert", "is_modular": true, "parents": [ "bert", "roberta" ], "root": "bert" }, "yolos": { "children": [], "depth": 0, "diff_size": 16, "extends": null, "is_modular": true, "parents": [], "root": "yolos" }, "youtu": { "children": [], "depth": 1, "diff_size": 31, "extends": "llama", "is_modular": true, "parents": [ "llama", "deepseek_v3", "qwen3" ], "root": "llama" }, "zamba": { "children": [ "zamba2" ], "depth": 0, "diff_size": null, "extends": null, "is_modular": false, "parents": [], "root": "zamba" }, "zamba2": { "children": [], "depth": 1, "diff_size": 44, "extends": "zamba", "is_modular": true, "parents": [ "zamba", "llama" ], "root": "zamba" } }, "provenance": { "generator": "utils/architecture_ir", "schema_version": "modular-graph-v0", "source": "ast_modular_files" }, "roots": [ "llama", "vit", "llava", "clip", "wav2vec2", "siglip", "bert", "detr", "qwen2_vl", "dinov2", "llava_next", "bart", "colpali", "pp_ocrv5_server_det", "sam", "chameleon", "depth_anything", "dpt", "esm", "fastspeech2_conformer", "gemma4_assistant", "granite_speech", "grounding_dino", "idefics3", "instructblip", "layoutlmv3", "mamba", "mllama", "moshi", "qwen2_audio", "resnet", "t5", "time_series_transformer", "xcodec", "zamba" ], "schema_version": "modular-graph-v0" }