Use HF Papers

candle-examples/examples/based/README.md

@@ -4,7 +4,7 @@ Experimental, not instruction-tuned small LLM from the Hazy Research group, comb
 
 [Blogpost](https://hazyresearch.stanford.edu/blog/2024-03-03-based)
 
-[Simple linear attention language models balance the recall-throughput tradeoff](https://arxiv.org/abs/2402.18668)
+[Simple linear attention language models balance the recall-throughput tradeoff](https://huggingface.co/papers/2402.18668)
 
 ## Running an example
 

candle-examples/examples/beit/README.md

@@ -1,6 +1,6 @@
 # candle-beit
 
-[Beit](https://arxiv.org/abs/2106.08254) is a computer vision model.
+[Beit](https://huggingface.co/papers/2106.08254) is a computer vision model.
 In this example, it is used as an ImageNet classifier: the model returns the
 probability for the image to belong to each of the 1000 ImageNet categories.
 

candle-examples/examples/colpali/README.md

@@ -3,7 +3,7 @@
 [HuggingFace Model Card](https://huggingface.co/vidore/colpali-v1.2-merged)
 
 ```
-wget https://arxiv.org/pdf/1706.03762.pdf
+wget https://huggingface.co/papers/1706.03762
 cargo run --features cuda,pdf2image --release --example colpali -- --prompt "What is Positional Encoding" --pdf "1706.03762.pdf"
 ```
 

candle-examples/examples/convmixer/README.md

@@ -2,7 +2,7 @@
 
 A lightweight CNN architecture that processes image patches similar to a vision transformer, with separate spatial and channel convolutions.
 
-ConvMixer from [Patches Are All You Need?](https://arxiv.org/pdf/2201.09792) and [ConvMixer](https://github.com/locuslab/convmixer). 
+ConvMixer from [Patches Are All You Need?](https://huggingface.co/papers/2201.09792) and [ConvMixer](https://github.com/locuslab/convmixer). 
 
 ## Running an example
 

candle-examples/examples/convnext/README.md

@@ -1,7 +1,7 @@
 # candle-convnext
 
-[A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) and
-[ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders](https://arxiv.org/abs/2301.00808).
+[A ConvNet for the 2020s](https://huggingface.co/papers/2201.03545) and
+[ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders](https://huggingface.co/papers/2301.00808).
 
 This candle implementation uses a pre-trained ConvNeXt network for inference. The
 classification head has been trained on the ImageNet dataset and returns the

candle-examples/examples/dinov2reg4/README.md

@@ -1,6 +1,6 @@
 # candle-dinov2-reg4
 
-[DINOv2-reg4](https://arxiv.org/abs/2309.16588) is the lastest version of DINOv2 with registers.
+[DINOv2-reg4](https://huggingface.co/papers/2309.16588) is the lastest version of DINOv2 with registers.
 In this example, it is used as an plant species classifier: the model returns the
 probability for the image to belong to each of the 7806 PlantCLEF2024 categories.
 

candle-examples/examples/dinov2reg4/main.rs

@@ -1,5 +1,5 @@
 //! DINOv2 reg4 finetuned on PlantCLEF 2024
-//! https://arxiv.org/abs/2309.16588
+//! https://huggingface.co/papers/2309.16588
 //! https://huggingface.co/spaces/BVRA/PlantCLEF2024
 //! https://zenodo.org/records/10848263
 

candle-examples/examples/efficientnet/main.rs

@@ -1,6 +1,6 @@
 //! EfficientNet implementation.
 //!
-//! https://arxiv.org/abs/1905.11946
+//! https://huggingface.co/papers/1905.11946
 
 #[cfg(feature = "mkl")]
 extern crate intel_mkl_src;

candle-examples/examples/efficientvit/README.md

@@ -1,6 +1,6 @@
 # candle-efficientvit
 
-[EfficientViT: Memory Efficient Vision Transformer with Cascaded Group Attention](https://arxiv.org/abs/2305.07027).
+[EfficientViT: Memory Efficient Vision Transformer with Cascaded Group Attention](https://huggingface.co/papers/2305.07027).
 
 This candle implementation uses a pre-trained EfficientViT (from Microsoft Research Asia) network for inference.
 The classification head has been trained on the ImageNet dataset and returns the probabilities for the top-5 classes.

candle-examples/examples/eva2/README.md

@@ -1,6 +1,6 @@
 # candle-eva2
 
-[EVA-02](https://arxiv.org/abs/2303.11331) is a computer vision model.
+[EVA-02](https://huggingface.co/papers/2303.11331) is a computer vision model.
 In this example, it is used as an ImageNet classifier: the model returns the
 probability for the image to belong to each of the 1000 ImageNet categories.
 

candle-examples/examples/fastvit/README.md

@@ -1,6 +1,6 @@
 # candle-fastvit
 
-[FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization](https://arxiv.org/abs/2303.14189).
+[FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization](https://huggingface.co/papers/2303.14189).
 This candle implementation uses a pre-trained FastViT network for inference. The
 classification head has been trained on the ImageNet dataset and returns the
 probabilities for the top-5 classes.

candle-examples/examples/gte-qwen/README.md

@@ -3,7 +3,7 @@
 gte-Qwen1.5-7B-instruct is a variant of the GTE embedding model family.
 
 - [Model card](https://huggingface.co/Alibaba-NLP/gte-Qwen1.5-7B-instruct) on the HuggingFace Hub.
-- [Technical report](https://arxiv.org/abs/2308.03281) *Towards General Text Embeddings with Multi-stage Contrastive Learning*
+- [Technical report](https://huggingface.co/papers/2308.03281) *Towards General Text Embeddings with Multi-stage Contrastive Learning*
 
 
 ## Running the example

candle-examples/examples/hiera/README.md

@@ -1,6 +1,6 @@
 # hiera
 
-[Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles](https://arxiv.org/abs/2306.00989)
+[Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles](https://huggingface.co/papers/2306.00989)
 This candle implementation uses pre-trained Hiera models from timm for inference.
 The classification head has been trained on the ImageNet dataset and returns the probabilities for the top-5 classes.
 

candle-examples/examples/mamba/README.md

@@ -5,7 +5,7 @@ the transformer architecture. It leverages State Space Models (SSMs) with the
 goal of being computationally efficient on long sequences. The implementation is
 based on [mamba.rs](https://github.com/LaurentMazare/mamba.rs).
 
-- [1]. [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://arxiv.org/abs/2312.00752).
+- [1]. [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://huggingface.co/papers/2312.00752).
 
 Compared to the mamba-minimal example, this version is far more efficient but
 would only work for inference.

candle-examples/examples/mobileclip/README.md

@@ -2,7 +2,7 @@
 
 MobileCLIP is family of efficient CLIP-like models using FastViT-based image encoders.
 
-See [MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training](https://arxiv.org/abs/2311.17049)
+See [MobileCLIP: Fast Image-Text Models through Multi-Modal Reinforced Training](https://huggingface.co/papers/2311.17049)
 
 
 ## Running on an example on cpu

candle-examples/examples/mobilenetv4/README.md

@@ -1,6 +1,6 @@
 # candle-mobilenetv4
 
-[MobileNetV4 - Universal Models for the Mobile Ecosystem](https://arxiv.org/abs/2404.10518)
+[MobileNetV4 - Universal Models for the Mobile Ecosystem](https://huggingface.co/papers/2404.10518)
 This candle implementation uses pre-trained MobileNetV4 models from timm for inference.
 The classification head has been trained on the ImageNet dataset and returns the probabilities for the top-5 classes.
 

candle-examples/examples/mobileone/README.md

@@ -1,6 +1,6 @@
 # candle-mobileone
 
-[MobileOne: An Improved One millisecond Mobile Backbone](https://arxiv.org/abs/2206.04040).
+[MobileOne: An Improved One millisecond Mobile Backbone](https://huggingface.co/papers/2206.04040).
 
 This candle implementation uses a pre-trained MobileOne network for inference. The
 classification head has been trained on the ImageNet dataset and returns the

candle-examples/examples/musicgen/README.md

@@ -1,6 +1,6 @@
 # candle-musicgen
 
-Candle implementation of musicgen from [Simple and Controllable Music Generation](https://arxiv.org/pdf/2306.05284).
+Candle implementation of musicgen from [Simple and Controllable Music Generation](https://huggingface.co/papers/2306.05284).
 
 ## Running an example
 

candle-examples/examples/olmo/README.md

@@ -3,7 +3,7 @@
 OLMo is a series of Open Language Models designed to enable the science of language models.
 
 - **Project Page:** https://allenai.org/olmo
-- **Papers:** [OLMo](https://arxiv.org/abs/2402.00838) [OLMo 2](https://arxiv.org/abs/2501.00656)
+- **Papers:** [OLMo](https://huggingface.co/papers/2402.00838) [OLMo 2](https://huggingface.co/papers/2501.00656)
 - **Technical blog post:** https://blog.allenai.org/olmo-open-language-model-87ccfc95f580
 - **W&B Logs:** https://wandb.ai/ai2-llm/OLMo-1B/reports/OLMo-1B--Vmlldzo2NzY1Njk1
 <!-- - **Press release:** TODO -->

candle-examples/examples/onnx/README.md

@@ -2,7 +2,7 @@
 
 This example demonstrates how to run [ONNX](https://github.com/onnx/onnx) based models in Candle.
 
-It contains small variants of two models, [SqueezeNet](https://arxiv.org/pdf/1602.07360.pdf) (default) and [EfficientNet](https://arxiv.org/pdf/1905.11946.pdf).
+It contains small variants of two models, [SqueezeNet](https://huggingface.co/papers/1602.07360) (default) and [EfficientNet](https://huggingface.co/papers/1905.11946).
 
 You can run the examples with following commands:
 

candle-examples/examples/quantized-t5/README.md

@@ -51,7 +51,7 @@ cargo run --example quantized-t5 --release  -- \
  Note that a storm surge is what forecasters consider a hurricane's most dangerous part.
 ```
 
-### [MADLAD-400](https://arxiv.org/abs/2309.04662)
+### [MADLAD-400](https://huggingface.co/papers/2309.04662)
 
 MADLAD-400 is a series of multilingual machine translation T5 models trained on 250 billion tokens covering over 450 languages using publicly available data. These models are competitive with significantly larger models.
 

candle-examples/examples/repvgg/README.md

@@ -1,6 +1,6 @@
 # candle-repvgg
 
-[RepVGG: Making VGG-style ConvNets Great Again](https://arxiv.org/abs/2101.03697).
+[RepVGG: Making VGG-style ConvNets Great Again](https://huggingface.co/papers/2101.03697).
 
 This candle implementation uses a pre-trained RepVGG network for inference. The
 classification head has been trained on the ImageNet dataset and returns the

candle-examples/examples/resnet/README.md

@@ -1,6 +1,6 @@
 # candle-resnet
 
-A candle implementation of inference using a pre-trained [ResNet](https://arxiv.org/abs/1512.03385).
+A candle implementation of inference using a pre-trained [ResNet](https://huggingface.co/papers/1512.03385).
 This uses a classification head trained on the ImageNet dataset and returns the
 probabilities for the top-5 classes.
 

candle-examples/examples/stable-diffusion-3/README.md

@@ -7,7 +7,7 @@
 Stable Diffusion 3 Medium is a text-to-image model based on Multimodal Diffusion Transformer (MMDiT) architecture.
 
 - [huggingface repo](https://huggingface.co/stabilityai/stable-diffusion-3-medium)
-- [research paper](https://arxiv.org/pdf/2403.03206)
+- [research paper](https://huggingface.co/papers/2403.03206)
 - [announcement blog post](https://stability.ai/news/stable-diffusion-3-medium)
 
 Stable Diffusion 3.5 is a family of text-to-image models with latest improvements:

candle-examples/examples/stable-diffusion-3/sampling.rs

@@ -69,7 +69,7 @@ pub fn euler_sample(
 }
 
 // The "Resolution-dependent shifting of timestep schedules" recommended in the SD3 tech report paper
-// https://arxiv.org/pdf/2403.03206
+// https://huggingface.co/papers/2403.03206
 // Following the implementation in ComfyUI:
 // https://github.com/comfyanonymous/ComfyUI/blob/3c60ecd7a83da43d694e26a77ca6b93106891251/
 // comfy/model_sampling.py#L181

candle-examples/examples/starcoder2/README.md

@@ -1,6 +1,6 @@
 # candle-starcoder2
 
-Candle implementation of Star Coder 2 family of code generation model from [StarCoder 2 and The Stack v2: The Next Generation](https://arxiv.org/pdf/2402.19173).
+Candle implementation of Star Coder 2 family of code generation model from [StarCoder 2 and The Stack v2: The Next Generation](https://huggingface.co/papers/2402.19173).
 
 ## Running an example
 

candle-examples/examples/stella-en-v5/README.md

@@ -16,7 +16,7 @@ $ cargo run --example stella-en-v5 --release  -- --query "What are safetensors?"
 >  Tensor[[1, 1024], f32]
 ```
 
-Stella_en_1.5B_v5 is trained by [MRL](https://arxiv.org/abs/2205.13147) enabling multiple embedding dimensions.
+Stella_en_1.5B_v5 is trained by [MRL](https://huggingface.co/papers/2205.13147) enabling multiple embedding dimensions.
 
 The following reproduces the example in the [model card](https://huggingface.co/dunzhang/stella_en_1.5B_v5) for a retrieval task (s2p). The sample queries and docs are hardcoded in the example.
 

candle-examples/examples/t5/README.md

@@ -13,7 +13,7 @@ $ cargo run --example t5 --release -- --model-id "t5-small" --prompt "translate
 
 Variants such as [flan-t5](https://huggingface.co/google/flan-t5-small), [flan-ul2](https://huggingface.co/google/flan-ul2) (with `--revision "refs/pr/25"`), and [Co-EdIT](https://huggingface.co/grammarly/coedit-large) are also supported.
 
-## Translation with [MADLAD-400](https://arxiv.org/abs/2309.04662)
+## Translation with [MADLAD-400](https://huggingface.co/papers/2309.04662)
 
 MADLAD-400 is a series of multilingual machine translation T5 models trained on 250 billion tokens covering over 450 languages using publicly available data. These models are competitive with significantly larger models.
 

candle-examples/examples/wuerstchen/README.md

@@ -8,7 +8,7 @@ The candle implementation reproduces the same structure/files for models and
 pipelines. Useful resources:
 
 - [Official implementation](https://github.com/dome272/Wuerstchen).
-- [Arxiv paper](https://arxiv.org/abs/2306.00637).
+- [Arxiv paper](https://huggingface.co/papers/2306.00637).
 - Blog post: [Introducing Würstchen: Fast Diffusion for Image Generation](https://huggingface.co/blog/wuerstchen).
 
 ## Getting the weights