Skip to content

feat: add WinML hub catalog and per-model precision configs #431

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
ssss141414 wants to merge 1 commit into from
Closed

feat: add WinML hub catalog and per-model precision configs #431

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
344 changes: 344 additions & 0 deletions .aitk/configs/wmk_hub_catalog.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,344 @@
[
{
"model_id": "BAAI/bge-base-en-v1.5",
"description": "General-purpose English text embedding model (110M params) for retrieval and ranking.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "BAAI/bge-base-en-v1.5",
"description": "General-purpose English embedding model optimized for semantic similarity.",
"model_type": "bert",
"task": "sentence-similarity"
},
{
"model_id": "BAAI/bge-large-en-v1.5",
"description": "High-capacity English text embedding model (335M params) for semantic similarity.",
"model_type": "bert",
"task": "sentence-similarity"
},
{
"model_id": "BAAI/bge-small-en-v1.5",
"description": "Compact English text embedding model (33M params) for retrieval and ranking.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "BAAI/bge-small-en-v1.5",
"description": "Compact English embedding model optimized for semantic similarity and matching.",
"model_type": "bert",
"task": "sentence-similarity"
},
{
"model_id": "Babelscape/wikineural-multilingual-ner",
"description": "Multilingual BERT model for named entity recognition across 9 languages.",
"model_type": "bert",
"task": "token-classification"
},
{
"model_id": "FacebookAI/roberta-base",
"description": "RoBERTa base — robustly optimized BERT pretraining for English masked language modeling.",
"model_type": "roberta",
"task": "fill-mask"
},
{
"model_id": "FacebookAI/roberta-large",
"description": "RoBERTa large (24 layers, 355M params) for English masked language modeling.",
"model_type": "roberta",
"task": "fill-mask"
},
{
"model_id": "FacebookAI/xlm-roberta-base",
"description": "Multilingual RoBERTa base trained on 100 languages for masked language modeling.",
"model_type": "xlm-roberta",
"task": "fill-mask"
},
{
"model_id": "FacebookAI/xlm-roberta-large",
"description": "Multilingual RoBERTa large (550M params) for cross-lingual masked language modeling.",
"model_type": "xlm-roberta",
"task": "fill-mask"
},
{
"model_id": "Intel/bert-base-uncased-mrpc",
"description": "BERT-base fine-tuned on MRPC paraphrase corpus for text embedding extraction.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "Intel/bert-base-uncased-mrpc",
"description": "BERT-base fine-tuned on MRPC for paraphrase detection and text classification.",
"model_type": "bert",
"task": "text-classification"
},
{
"model_id": "ProsusAI/finbert",
"description": "Financial sentiment analysis model built on BERT, classifying text as positive, negative, or neutral.",
"model_type": "bert",
"task": "text-classification"
},
{
"model_id": "StanfordAIMI/dinov2-base-xray-224",
"description": "DINOv2 base fine-tuned on chest X-rays for medical image feature extraction.",
"model_type": "dinov2",
"task": "image-feature-extraction"
},
{
"model_id": "cardiffnlp/twitter-roberta-base-sentiment-latest",
"description": "RoBERTa model fine-tuned on ~124M tweets for sentiment analysis (positive/negative/neutral).",
"model_type": "roberta",
"task": "text-classification"
},
{
"model_id": "dbmdz/bert-large-cased-finetuned-conll03-english",
"description": "BERT-large model fine-tuned on CoNLL-2003 for English named entity recognition.",
"model_type": "bert",
"task": "token-classification"
},
{
"model_id": "deepset/bert-large-uncased-whole-word-masking-squad2",
"description": "BERT-large with whole word masking, fine-tuned on SQuAD 2.0 for question answering.",
"model_type": "bert",
"task": "question-answering"
},
{
"model_id": "deepset/roberta-base-squad2",
"description": "RoBERTa-base fine-tuned on SQuAD 2.0 for extractive question answering.",
"model_type": "roberta",
"task": "question-answering"
},
{
"model_id": "deepset/tinyroberta-squad2",
"description": "Compact RoBERTa model fine-tuned on SQuAD 2.0 for lightweight question answering.",
"model_type": "roberta",
"task": "question-answering"
},
{
"model_id": "dslim/bert-base-NER",
"description": "BERT model fine-tuned on CoNLL-2003 for named entity recognition (PER, ORG, LOC, MISC).",
"model_type": "bert",
"task": "token-classification"
},
{
"model_id": "facebook/convnext-tiny-224",
"description": "ConvNeXt-Tiny model combining CNN efficiency with Transformer-era design for image classification.",
"model_type": "convnext",
"task": "image-classification"
},
{
"model_id": "facebook/dino-vitb16",
"description": "Vision Transformer base (ViT-B/16) self-supervised with DINO for image feature extraction.",
"model_type": "vit",
"task": "image-feature-extraction"
},
{
"model_id": "facebook/dino-vits16",
"description": "Vision Transformer small (ViT-S/16) self-supervised with DINO for image feature extraction.",
"model_type": "vit",
"task": "image-feature-extraction"
},
{
"model_id": "facebook/dinov2-base",
"description": "DINOv2 base self-supervised vision model for general-purpose image feature extraction.",
"model_type": "dinov2",
"task": "image-feature-extraction"
},
{
"model_id": "facebook/dinov2-large",
"description": "DINOv2 large (300M params) self-supervised vision model for image feature extraction.",
"model_type": "dinov2",
"task": "image-feature-extraction"
},
{
"model_id": "facebook/dinov2-small",
"description": "DINOv2 small self-supervised vision model for efficient image feature extraction.",
"model_type": "dinov2",
"task": "image-feature-extraction"
},
{
"model_id": "google-bert/bert-base-multilingual-cased",
"description": "Multilingual BERT (104 languages) for general-purpose text embeddings.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "google-bert/bert-base-multilingual-uncased",
"description": "Multilingual BERT base (uncased) pretrained on 102 languages for masked language modeling.",
"model_type": "bert",
"task": "fill-mask"
},
{
"model_id": "google-bert/bert-base-uncased",
"description": "BERT base uncased pretrained on English text for masked language modeling.",
"model_type": "bert",
"task": "fill-mask"
},
{
"model_id": "google-bert/bert-large-uncased-whole-word-masking-finetuned-squad",
"description": "BERT-large with whole word masking, fine-tuned on SQuAD for question answering.",
"model_type": "bert",
"task": "question-answering"
},
{
"model_id": "google/vit-base-patch16-224",
"description": "Vision Transformer (ViT) pre-trained on ImageNet-21k, fine-tuned on ImageNet-1k at 224x224.",
"model_type": "vit",
"task": "image-classification"
},
{
"model_id": "google/vit-base-patch16-224-in21k",
"description": "Vision Transformer base pretrained on ImageNet-21k for image feature extraction.",
"model_type": "vit",
"task": "image-feature-extraction"
},
{
"model_id": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
"description": "LAION CLIP ViT-B/32 trained on 2B image-text pairs for joint image/text feature extraction.",
"model_type": "clip",
"task": "feature-extraction"
},
{
"model_id": "laion/CLIP-ViT-B-32-laion2B-s34B-b79K",
"description": "LAION CLIP ViT-B/32 for zero-shot image classification via image-text similarity.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "laion/CLIP-ViT-H-14-laion2B-s32B-b79K",
"description": "LAION CLIP ViT-H/14 large model for high-accuracy zero-shot image classification.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "mattmdjaga/segformer_b2_clothes",
"description": "SegFormer-B2 fine-tuned for clothing segmentation in fashion images.",
"model_type": "segformer",
"task": "image-segmentation"
},
{
"model_id": "microsoft/rad-dino",
"description": "DINOv2-based vision model fine-tuned on chest X-rays for radiology feature extraction.",
"model_type": "dinov2",
"task": "image-feature-extraction"
},
{
"model_id": "microsoft/resnet-50",
"description": "Classic ResNet-50 model pre-trained on ImageNet-1k for image classification.",
"model_type": "resnet",
"task": "image-classification"
},
{
"model_id": "microsoft/swin-large-patch4-window7-224",
"description": "Swin Transformer large model for image classification at 224x224 resolution.",
"model_type": "swin",
"task": "image-classification"
},
{
"model_id": "microsoft/table-transformer-detection",
"description": "DETR-based model for detecting tables in document images.",
"model_type": "table-transformer",
"task": "object-detection"
},
{
"model_id": "nvidia/segformer-b1-finetuned-ade-512-512",
"description": "SegFormer-B1 fine-tuned on ADE20K for semantic segmentation at 512x512.",
"model_type": "segformer",
"task": "image-segmentation"
},
{
"model_id": "nvidia/segformer-b2-finetuned-ade-512-512",
"description": "SegFormer-B2 fine-tuned on ADE20K for semantic segmentation at 512x512.",
"model_type": "segformer",
"task": "image-segmentation"
},
{
"model_id": "nvidia/segformer-b5-finetuned-ade-640-640",
"description": "SegFormer-B5 (largest variant) fine-tuned on ADE20K for semantic segmentation at 640x640.",
"model_type": "segformer",
"task": "image-segmentation"
},
{
"model_id": "openai/clip-vit-base-patch16",
"description": "CLIP ViT-B/16 model for joint image-text embeddings with 16x16 patch size.",
"model_type": "clip",
"task": "feature-extraction"
},
{
"model_id": "openai/clip-vit-base-patch16",
"description": "OpenAI CLIP ViT-B/16 for zero-shot image classification via image-text similarity.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "openai/clip-vit-base-patch32",
"description": "CLIP ViT-B/32 model for joint image-text embeddings with 32x32 patch size.",
"model_type": "clip",
"task": "feature-extraction"
},
{
"model_id": "openai/clip-vit-base-patch32",
"description": "OpenAI CLIP ViT-B/32 for zero-shot image classification via image-text similarity.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "openai/clip-vit-large-patch14",
"description": "OpenAI CLIP ViT-L/14 for high-accuracy zero-shot image classification.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "openai/clip-vit-large-patch14-336",
"description": "OpenAI CLIP ViT-L/14 at 336px resolution for higher-accuracy zero-shot image classification.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "patrickjohncyh/fashion-clip",
"description": "CLIP fine-tuned on fashion product images for fashion-specific zero-shot classification.",
"model_type": "clip",
"task": "zero-shot-image-classification"
},
{
"model_id": "rizvandwiki/gender-classification",
"description": "ViT model fine-tuned for gender classification from facial images.",
"model_type": "vit",
"task": "image-classification"
},
{
"model_id": "sentence-transformers/all-MiniLM-L6-v2",
"description": "Lightweight sentence embedding model mapping text to 384-dim dense vectors.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "sentence-transformers/all-MiniLM-L6-v2",
"description": "Lightweight sentence embedding model optimized for semantic similarity tasks.",
"model_type": "bert",
"task": "sentence-similarity"
},
{
"model_id": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
"description": "Multilingual sentence embedding model supporting 50+ languages, 384-dim output.",
"model_type": "bert",
"task": "feature-extraction"
},
{
"model_id": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
"description": "Multilingual model for cross-lingual semantic similarity across 50+ languages.",
"model_type": "bert",
"task": "sentence-similarity"
},
{
"model_id": "sentence-transformers/paraphrase-multilingual-mpnet-base-v2",
"description": "Multilingual MPNet model for high-quality semantic similarity across 50+ languages.",
"model_type": "xlm-roberta",
"task": "sentence-similarity"
},
{
"model_id": "w11wo/indonesian-roberta-base-posp-tagger",
"description": "RoBERTa model fine-tuned for Indonesian part-of-speech tagging.",
"model_type": "roberta",
"task": "token-classification"
}
]
Loading
Loading