Use HF Papers

candle-transformers/src/models/based.rs

@@ -1,7 +1,7 @@
 //! Based from the Stanford Hazy Research group.
 //!
 //! See "Simple linear attention language models balance the recall-throughput tradeoff", Arora et al. 2024
-//! - Simple linear attention language models balance the recall-throughput tradeoff. [Arxiv](https://arxiv.org/abs/2402.18668)
+//! - Simple linear attention language models balance the recall-throughput tradeoff. [Arxiv](https://huggingface.co/papers/2402.18668)
 //! - [Github Rep](https://github.com/HazyResearch/based)
 //! - [Blogpost](https://hazyresearch.stanford.edu/blog/2024-03-03-based)
 

candle-transformers/src/models/beit.rs

@@ -1,7 +1,7 @@
 //! Based on the BEIT vision-language model.
 //!
 //! See "BEIT: BERT Pre-Training of Image Transformers", Bao et al. 2021
-//! - [Arxiv](https://arxiv.org/abs/2106.08254)
+//! - [Arxiv](https://huggingface.co/papers/2106.08254)
 //! - [Github](https://github.com/microsoft/unilm/tree/master/beit)
 //!
 

candle-transformers/src/models/bert.rs

@@ -3,7 +3,7 @@
 //! Bert is a general large language model that can be used for various language tasks:
 //! - Compute sentence embeddings for a prompt.
 //! - Compute similarities between a set of sentences.
-//! - [Arxiv](https://arxiv.org/abs/1810.04805) "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding"
+//! - [Arxiv](https://huggingface.co/papers/1810.04805) "BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding"
 //! - Upstream [Github repo](https://github.com/google-research/bert).
 //! - See bert in [candle-examples](https://github.com/huggingface/candle/tree/main/candle-examples/) for runnable code
 //!

candle-transformers/src/models/bigcode.rs

@@ -3,7 +3,7 @@
 //! [StarCoder/BigCode](https://huggingface.co/bigcode/starcoderbase-1b) is a LLM
 //! model specialized to code generation. The initial model was trained on 80
 //! programming languages. See "StarCoder: A State-of-the-Art LLM for Code", Mukherjee et al. 2023
-//! - [Arxiv](https://arxiv.org/abs/2305.06161)
+//! - [Arxiv](https://huggingface.co/papers/2305.06161)
 //! - [Github](https://github.com/bigcode-project/starcoder)
 //!
 //! ## Running some example

candle-transformers/src/models/blip.rs

@@ -5,7 +5,7 @@
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/radames/Candle-BLIP-Image-Captioning)
 //! - 💻 [GH Link](https://github.com/salesforce/BLIP)
 //! - 🤗 [HF Link](https://huggingface.co/Salesforce/blip-image-captioning-base)
-//! - 📝 [Paper](https://arxiv.org/abs/2201.12086)
+//! - 📝 [Paper](https://huggingface.co/papers/2201.12086)
 //!
 
 use super::blip_text;

candle-transformers/src/models/blip_text.rs

@@ -1,11 +1,11 @@
 //! Implementation of BLIP text encoder/decoder.
 //!
-//! - 📝 [Paper](https://arxiv.org/abs/2201.12086). BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation"
+//! - 📝 [Paper](https://huggingface.co/papers/2201.12086). BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation"
 //!
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/radames/Candle-BLIP-Image-Captioning)
 //! - 💻 [GH Link](https://github.com/salesforce/BLIP)
 //! - 🤗 [HF Link](https://huggingface.co/Salesforce/blip-image-captioning-base)
-//! - 📝 [Paper](https://arxiv.org/abs/2201.12086)
+//! - 📝 [Paper](https://huggingface.co/papers/2201.12086)
 //!
 use super::with_tracing::{linear, Embedding, Linear};
 use candle::{Module, Result, Tensor, D};

candle-transformers/src/models/chinese_clip/text_model.rs

@@ -13,9 +13,9 @@ use super::Activation;
 
 /// Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query"`. For
 /// positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to
-/// [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155).
+/// [Self-Attention with Relative Position Representations (Shaw et al.)](https://huggingface.co/papers/1803.02155).
 /// For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models
-/// with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658).
+/// with Better Relative Position Embeddings (Huang et al.)](https://huggingface.co/papers/2009.13658).
 #[derive(Clone, Debug)]
 pub enum PositionEmbeddingType {
     Absolute,

candle-transformers/src/models/codegeex4_9b.rs

@@ -2,7 +2,7 @@
 //!
 //! A Pre-Trained Model For Code Generation with Multilingual Evaluations on HumanEval-X"
 //!
-//! - 📝 [Arxiv](https://arxiv.org/abs/2303.17568)
+//! - 📝 [Arxiv](https://huggingface.co/papers/2303.17568)
 //! - 💻 [Github](https://github.com/THUDM/CodeGeeX)
 //!
 

candle-transformers/src/models/convmixer.rs

@@ -2,7 +2,7 @@
 //!
 //! See "Patches Are All You Need?" by Trockman et al. 2022
 //!
-//! - 📝 [Arxiv](https://arxiv.org/abs/2201.09792)
+//! - 📝 [Arxiv](https://huggingface.co/papers/2201.09792)
 //! - 💻 [Github](https://github.com/locuslab/convmixer)
 //!
 use candle::Result;

candle-transformers/src/models/convnext.rs

@@ -8,8 +8,8 @@
 //! - 💻 [ConvNeXt](https://github.com/facebookresearch/ConvNeXt/)
 //! - 💻 [ConvNeXt-V2](https://github.com/facebookresearch/ConvNeXt-V2/)
 //! - 💻 [timm](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/convnext.py)
-//! - 📝 [Paper](https://arxiv.org/abs/2201.03545) A ConvNet for the 2020s
-//! - 📝 [Paper](https://arxiv.org/abs/2301.00808) ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders
+//! - 📝 [Paper](https://huggingface.co/papers/2201.03545) A ConvNet for the 2020s
+//! - 📝 [Paper](https://huggingface.co/papers/2301.00808) ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders
 //!
 
 use candle::shape::ShapeWithOneHole;

candle-transformers/src/models/dinov2reg4.rs

@@ -4,7 +4,7 @@
 //! original architecture. This implementation is specifically trained for plant species
 //! classification on the PlantCLEF2024 dataset with 7,806 classes.
 //!
-//! - [Paper](https://arxiv.org/abs/2309.16588). DINOv2: Learning Robust Visual Features without Supervision
+//! - [Paper](https://huggingface.co/papers/2309.16588). DINOv2: Learning Robust Visual Features without Supervision
 //! - [GH Repo](https://github.com/facebookresearch/dinov2)
 //!
 //! # Example

candle-transformers/src/models/distilbert.rs

@@ -1,7 +1,7 @@
 //! Implementation of DistilBert, a distilled version of BERT.
 //!
 //! See:
-//! - ["DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter"](https://arxiv.org/abs/1910.01108)
+//! - ["DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter"](https://huggingface.co/papers/1910.01108)
 //!
 use super::with_tracing::{layer_norm, linear, LayerNorm, Linear};
 use candle::{DType, Device, Result, Tensor};

candle-transformers/src/models/efficientnet.rs

@@ -1,7 +1,7 @@
 //! Implementation of EfficientBert, an efficient variant of BERT for computer vision tasks.
 //!
 //! See:
-//! - ["EfficientBERT: Progressively Searching Multilayer Perceptron Architectures for BERT"](https://arxiv.org/abs/2201.00462)
+//! - ["EfficientBERT: Progressively Searching Multilayer Perceptron Architectures for BERT"](https://huggingface.co/papers/2201.00462)
 //!
 use candle::{Context, Result, Tensor, D};
 use candle_nn as nn;

candle-transformers/src/models/efficientvit.rs

@@ -5,7 +5,7 @@
 //! to achieve strong performance while maintaining low memory usage.
 //!
 //! The model was originally described in the paper:
-//! ["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 implementation is based on the reference implementation from
 //! [pytorch-image-models](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/efficientvit_msra.py).

candle-transformers/src/models/encodec.rs

@@ -1,6 +1,6 @@
 //! EnCodec neural audio codec based on the Encodec implementation.
 //!
-//! See ["High Fidelity Neural Audio Compression"](https://arxiv.org/abs/2210.13438)
+//! See ["High Fidelity Neural Audio Compression"](https://huggingface.co/papers/2210.13438)
 //!
 //! Based on implementation from [huggingface/transformers](https://github.com/huggingface/transformers/blob/main/src/transformers/models/encodec/modeling_encodec.py)
 

candle-transformers/src/models/eva2.rs

@@ -4,7 +4,7 @@
 //! The model returns the probability for an image to belong to each of the 1000
 //! ImageNet categories.
 //!
-//! - [Paper](https://arxiv.org/abs/2303.11331). EVA-02: A Visual Representation for Neon Genesis
+//! - [Paper](https://huggingface.co/papers/2303.11331). EVA-02: A Visual Representation for Neon Genesis
 //! - [Code](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/eva2.py)
 //!
 //! # Example

candle-transformers/src/models/fastvit.rs

@@ -1,7 +1,7 @@
 //! # FastViT inference implementation based on timm
 //!
 //! ## Description
-//! See ["FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization"](https://arxiv.org/pdf/2303.14189)
+//! See ["FastViT: A Fast Hybrid Vision Transformer using Structural Reparameterization"](https://huggingface.co/papers/2303.14189)
 //!
 //! Implementation based on [timm model](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/fastvit.py)
 

candle-transformers/src/models/hiera.rs

@@ -2,7 +2,7 @@
 //!
 //!
 //! - 💻 [Hiera](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/hiera.py)
-//! - 📝 [Paper](https://arxiv.org/abs/2306.00989). Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles
+//! - 📝 [Paper](https://huggingface.co/papers/2306.00989). Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles
 
 use candle::{Result, D};
 use candle_nn::{conv2d, layer_norm, linear, ops::softmax, Conv2dConfig, Func, VarBuilder};

candle-transformers/src/models/llama.rs

@@ -1,6 +1,6 @@
 //! Llama inference implementation.
 //!
-//! See ["LLaMA: Open and Efficient Foundation Language Models"](https://arxiv.org/abs/2302.13971)
+//! See ["LLaMA: Open and Efficient Foundation Language Models"](https://huggingface.co/papers/2302.13971)
 //!
 //! Implementation based on Hugging Face's [transformers](https://github.com/huggingface/transformers/blob/main/src/transformers/models/llama/modeling_llama.py)
 

candle-transformers/src/models/llama2_c.rs

@@ -1,6 +1,6 @@
 //! Llama2 inference implementation.
 //!
-//! See ["LLaMA 2: Open Foundation and Fine-Tuned Chat Models"](https://arxiv.org/abs/2307.09288)
+//! See ["LLaMA 2: Open Foundation and Fine-Tuned Chat Models"](https://huggingface.co/papers/2307.09288)
 //!
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/lmz/candle-llama2)
 //! - 💻 llama2.c [GH Link](https://github.com/karpathy/llama2.c)

candle-transformers/src/models/llama2_c_weights.rs

@@ -1,6 +1,6 @@
 //! Llama2 inference implementation.
 //!
-//! See ["LLaMA 2: Open Foundation and Fine-Tuned Chat Models"](https://arxiv.org/abs/2307.09288)
+//! See ["LLaMA 2: Open Foundation and Fine-Tuned Chat Models"](https://huggingface.co/papers/2307.09288)
 //!
 //! Based on the [llama2.c](https://github.com/karpathy/llama2.c) implementation
 

candle-transformers/src/models/llava/mod.rs

@@ -4,7 +4,7 @@
 //! language model (Llama) for multimodal capabilities. The architecture implements the training-free projection technique.
 //!
 //! - 💻[GH Link](https://github.com/haotian-liu/LLaVA/tree/main)
-//! - 📝 [Paper](https://arxiv.org/abs/2304.08485)/ Visual Instruction Tuning
+//! - 📝 [Paper](https://huggingface.co/papers/2304.08485)/ Visual Instruction Tuning
 //!
 
 pub mod config;

candle-transformers/src/models/mamba.rs

@@ -1,6 +1,6 @@
 //! Mamba inference implementation.
 //!
-//! See ["Mamba: Linear-Time Sequence Modeling with Selective State Spaces"](https://arxiv.org/abs/2312.00752)
+//! See ["Mamba: Linear-Time Sequence Modeling with Selective State Spaces"](https://huggingface.co/papers/2312.00752)
 //!
 //! Based on reference implementation from the AlbertMamba project
 //! A fast implementation of mamba for inference only.
@@ -122,7 +122,7 @@ impl MambaBlock {
         let proj_for_conv = candle_nn::ops::silu(&proj_for_conv)?;
         // SSM + Selection, we're doing inference here so only need the last step of
         // the sequence.
-        // Algorithm 3.2 on page 6, https://arxiv.org/pdf/2312.00752.pdf
+        // Algorithm 3.2 on page 6, https://huggingface.co/papers/2312.00752
 
         let x_proj = self.x_proj.forward(&proj_for_conv)?;
         let delta = x_proj.narrow(D::Minus1, 0, self.dt_rank)?.contiguous()?;

candle-transformers/src/models/mixformer.rs

@@ -1,14 +1,14 @@
 //! MixFormer (Microsoft's Phi Architecture)
 //!
 //! See "Textbooks Are All You Need II: phi-1.5 technical report", Lin et al. 2023
-//! - [Arxiv](https://arxiv.org/abs/2309.05463)
+//! - [Arxiv](https://huggingface.co/papers/2309.05463)
 //! - [Github](https://huggingface.co/microsoft/phi-1_5)
 //!
 
 use crate::models::with_tracing::{linear, Embedding as E, Linear};
 /// MixFormer model.
 /// https://huggingface.co/microsoft/phi-1_5
-/// https://arxiv.org/abs/2309.05463
+/// https://huggingface.co/papers/2309.05463
 use candle::{DType, Device, IndexOp, Module, Result, Tensor, D};
 use candle_nn::{Activation, VarBuilder};
 use serde::Deserialize;

candle-transformers/src/models/mmdit/mod.rs

@@ -3,14 +3,14 @@
 //! Mix of Multi-scale Dilated and Traditional Convolutions (MMDiT) is an architecture
 //! introduced for Stable Diffusion 3, with the MMDiT-X variant used in Stable Diffusion 3.5.
 //!
-//! - 📝 [Research Paper](https://arxiv.org/abs/2403.03206)
+//! - 📝 [Research Paper](https://huggingface.co/papers/2403.03206)
 //! - 💻 ComfyUI [reference implementation](https://github.com/comfyanonymous/ComfyUI/blob/78e133d0415784924cd2674e2ee48f3eeca8a2aa/comfy/ldm/modules/diffusionmodules/mmdit.py)
 //! - 💻 Stability-AI [MMDiT-X implementation](https://github.com/Stability-AI/sd3.5/blob/4e484e05308d83fb77ae6f680028e6c313f9da54/mmditx.py)
 
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/radames/Candle-BLIP-Image-Captioning)
 //! - 💻 [GH Link](https://github.com/salesforce/BLIP)
 //! - 🤗 [HF Link](https://huggingface.co/Salesforce/blip-image-captioning-base)
-//! - 📝 [Paper](https://arxiv.org/abs/2201.12086)
+//! - 📝 [Paper](https://huggingface.co/papers/2201.12086)
 //!
 
 pub mod blocks;

candle-transformers/src/models/mmdit/model.rs

@@ -1,4 +1,4 @@
-// Implement the MMDiT model originally introduced for Stable Diffusion 3 (https://arxiv.org/abs/2403.03206),
+// Implement the MMDiT model originally introduced for Stable Diffusion 3 (https://huggingface.co/papers/2403.03206),
 // as well as the MMDiT-X variant introduced for Stable Diffusion 3.5-medium (https://huggingface.co/stabilityai/stable-diffusion-3.5-medium)
 // This follows the implementation of the MMDiT model in the ComfyUI repository.
 // https://github.com/comfyanonymous/ComfyUI/blob/78e133d0415784924cd2674e2ee48f3eeca8a2aa/comfy/ldm/modules/diffusionmodules/mmdit.py#L1

candle-transformers/src/models/mobileclip.rs

@@ -6,12 +6,12 @@
 //! - Projection layers to align the feature spaces
 //!
 //! See model details at:
-//! - [FastViT](https://arxiv.org/abs/2303.14189)
+//! - [FastViT](https://huggingface.co/papers/2303.14189)
 //! - [OpenCLIP](https://github.com/mlfoundations/open_clip)
 //!
 //! References:
 //! - [MobileVLM](https://huggingface.co/mobileVLM)
-//! - [MetaCLIP](https://arxiv.org/abs/2309.16671)
+//! - [MetaCLIP](https://huggingface.co/papers/2309.16671)
 //!
 
 use super::fastvit;

candle-transformers/src/models/mobilenetv4.rs

@@ -4,7 +4,7 @@
 //!
 //! ## Paper
 //!
-//! ["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)
 //!
 //! ## References
 //!

candle-transformers/src/models/mobileone.rs

@@ -2,7 +2,7 @@
 //!
 //! MobileOne inference implementation based on timm and candle-repvgg
 //!
-//! See ["MobileOne: An Improved One millisecond Mobile Backbone"](https://arxiv.org/abs/2206.04040)
+//! See ["MobileOne: An Improved One millisecond Mobile Backbone"](https://huggingface.co/papers/2206.04040)
 
 use candle::{DType, Result, Tensor, D};
 use candle_nn::{

candle-transformers/src/models/modernbert.rs

@@ -1,7 +1,7 @@
 //! ModernBERT
 //!
 //! ModernBERT is a modernized bidirectional encoder-only Transformer model.
-//! - [Arxiv](https://arxiv.org/abs/2412.13663) "Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference"
+//! - [Arxiv](https://huggingface.co/papers/2412.13663) "Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference"
 //! - Upstream [Github repo](https://github.com/AnswerDotAI/ModernBERT).
 //! - See modernbert in [candle-examples](https://github.com/huggingface/candle/tree/main/candle-examples/) for runnable code
 //!

candle-transformers/src/models/nvembed_v2/mod.rs

@@ -2,7 +2,7 @@
 //!
 //! NV-Embed-v2 is a text embedding model that combines a Mistral decoder with a latent attention mechanism to produce high-quality text embeddings.
 //!
-//! This implementation is based on the [paper](https://arxiv.org/pdf/2405.17428) and [weights](https://huggingface.co/nvidia/NV-Embed-v2)
+//! This implementation is based on the [paper](https://huggingface.co/papers/2405.17428) and [weights](https://huggingface.co/nvidia/NV-Embed-v2)
 //!
 //! # Query-Passage Retrieval Example
 //! ```bash

candle-transformers/src/models/olmo2.rs

@@ -2,7 +2,7 @@
 //!
 //! See OLMo 2 model details at:
 //! - [Hugging Face Collection](https://huggingface.co/collections/allenai/olmo-2-674117b93ab84e98afc72edc)
-//! - [OLMo 2 Paper](https://arxiv.org/abs/2501.00656)
+//! - [OLMo 2 Paper](https://huggingface.co/papers/2501.00656)
 //!
 //!
 use candle::{DType, Device, Module, Result, Tensor, D};

candle-transformers/src/models/openclip/mod.rs

@@ -4,7 +4,7 @@
 //! pairs of images with related texts.
 //!
 //! - 💻 [GH Link](https://github.com/mlfoundations/open_clip)
-//! - 📝 [Paper](https://arxiv.org/abs/2212.07143)
+//! - 📝 [Paper](https://huggingface.co/papers/2212.07143)
 //!
 //! ## Overview
 //!

candle-transformers/src/models/paligemma.rs

@@ -1,7 +1,7 @@
 //! Multimodal multi-purpose model combining Gemma-based language model with SigLIP image understanding
 //!
 //! See PaLiGemma details at:
-//! - [Paper](https://arxiv.org/abs/2402.05257)
+//! - [Paper](https://huggingface.co/papers/2402.05257)
 //! - [Google Blog Post](https://blog.research.google/2024/02/paligemma-scaling-language-image.html)
 //!
 //! The model is a multimodal combination of:
@@ -11,7 +11,7 @@
 //!
 //! References:
 //! - [HuggingFace Implementation](https://huggingface.co/google/paligemma-3b)
-//! - [Paper: PaLI-3 and Beyond: Scaling Language-Image Learning](https://arxiv.org/abs/2402.05257)
+//! - [Paper: PaLI-3 and Beyond: Scaling Language-Image Learning](https://huggingface.co/papers/2402.05257)
 //!
 
 use crate::models::{gemma, siglip};

candle-transformers/src/models/quantized_blip.rs

@@ -10,7 +10,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - [BLIP Paper](https://arxiv.org/abs/2201.12086)
+//! - [BLIP Paper](https://huggingface.co/papers/2201.12086)
 //! - [Hugging Face Implementation](https://huggingface.co/docs/transformers/model_doc/blip)
 //!
 

candle-transformers/src/models/quantized_blip_text.rs

@@ -11,7 +11,7 @@
 //! - Quantized linear transformations
 //!
 //! References:
-//! - [BLIP Paper](https://arxiv.org/abs/2201.12086)
+//! - [BLIP Paper](https://huggingface.co/papers/2201.12086)
 //! - [Hugging Face Implementation](https://huggingface.co/docs/transformers/model_doc/blip)
 //!
 

candle-transformers/src/models/quantized_llama.rs

@@ -11,7 +11,7 @@
 //! - Configurable model sizes and parameter counts
 //!
 //! - 💻 [GH Link](https://github.com/facebookresearch/llama)
-//! - 📝 [Paper](https://arxiv.org/abs/2302.13971)
+//! - 📝 [Paper](https://huggingface.co/papers/2302.13971)
 //!
 //! ![](https://raw.githubusercontent.com/huggingface/candle/main/candle-examples/examples/quantized/assets/aoc.gif)
 //!

candle-transformers/src/models/quantized_llama2_c.rs

@@ -10,7 +10,7 @@
 //! - 8-bit quantization of weights
 //!
 //! References:
-//! - [LLaMA2 Paper](https://arxiv.org/abs/2307.09288)
+//! - [LLaMA2 Paper](https://huggingface.co/papers/2307.09288)
 //! - [LLaMA2 Technical Report](https://ai.meta.com/research/publications/llama-2-open-foundation-and-fine-tuned-chat-models/)
 //!
 

candle-transformers/src/models/quantized_mistral.rs

@@ -11,7 +11,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - [Mistral Paper](https://arxiv.org/abs/2310.06825)
+//! - [Mistral Paper](https://huggingface.co/papers/2310.06825)
 //! - [Model Card](https://huggingface.co/mistralai/Mistral-7B-v0.1)
 //!
 

candle-transformers/src/models/quantized_phi.rs

@@ -11,7 +11,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - [Phi2 Paper](https://arxiv.org/abs/2309.05463)
+//! - [Phi2 Paper](https://huggingface.co/papers/2309.05463)
 //! - [Model Card](https://huggingface.co/microsoft/phi-2)
 //!
 

candle-transformers/src/models/quantized_recurrent_gemma.rs

@@ -11,7 +11,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - [Gemma Paper](https://arxiv.org/abs/2401.06751)
+//! - [Gemma Paper](https://huggingface.co/papers/2401.06751)
 //! - [Model Card](https://ai.google.dev/gemma)
 //!
 

candle-transformers/src/models/quantized_t5.rs

@@ -11,7 +11,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - 📝 [T5 Paper](https://arxiv.org/abs/1910.10683)
+//! - 📝 [T5 Paper](https://huggingface.co/papers/1910.10683)
 //! - 🤗 [Model Card](https://huggingface.co/t5-base)
 //! - 🤗 Original model from [T5](https://github.com/huggingface/transformers/blob/main/src/transformers/models/t5/modeling_t5.py)
 

candle-transformers/src/models/qwen2_moe.rs

@@ -12,7 +12,7 @@
 //! - Rotary positional embeddings (RoPE)
 //!
 //! References:
-//! - [Qwen2 Paper](https://arxiv.org/abs/2401.08985)
+//! - [Qwen2 Paper](https://huggingface.co/papers/2401.08985)
 //! - [Model Card](https://huggingface.co/Qwen/Qwen2-7B-beta)
 //!
 

candle-transformers/src/models/recurrent_gemma.rs

@@ -12,7 +12,7 @@
 //!
 //! References:
 //! - [Gemma: Open Models Based on Gemini Technology](https://blog.google/technology/developers/gemma-open-models/)
-//! - [Recurrent Memory model architecture](https://arxiv.org/abs/2402.00441)
+//! - [Recurrent Memory model architecture](https://huggingface.co/papers/2402.00441)
 //!
 //! This implementation is based on the python version from huggingface/transformers.
 //! https://github.com/huggingface/transformers/blob/b109257f4fb8b1166e7c53cc5418632014ed53a5/src/transformers/models/recurrent_gemma/modeling_recurrent_gemma.py#L2

candle-transformers/src/models/repvgg.rs

@@ -7,7 +7,7 @@
 //! - High accuracy with VGG-like plain architecture and training
 //!
 //! References:
-//! - [RepVGG Paper](https://arxiv.org/abs/2101.03697). RepVGG: Making VGG-style ConvNets Great Again
+//! - [RepVGG Paper](https://huggingface.co/papers/2101.03697). RepVGG: Making VGG-style ConvNets Great Again
 //! - [Official Implementation](https://github.com/DingXiaoH/RepVGG)
 //!
 

candle-transformers/src/models/resnet.rs

@@ -4,7 +4,7 @@
 //!
 //! ## Reference
 //!
-//! [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385)
+//! [Deep Residual Learning for Image Recognition](https://huggingface.co/papers/1512.03385)
 //! He et al. (2015)
 //!
 //! This paper introduced ResNet, a deep neural network architecture that utilizes

candle-transformers/src/models/segformer.rs

@@ -10,7 +10,7 @@
 //! - Lightweight all-MLP decode head
 //!
 //! References:
-//! - [SegFormer Paper](https://arxiv.org/abs/2105.15203)
+//! - [SegFormer Paper](https://huggingface.co/papers/2105.15203)
 //! - [Model Card](https://huggingface.co/nvidia/mit-b0)
 //!
 

candle-transformers/src/models/segment_anything/mod.rs

@@ -8,7 +8,7 @@
 //!
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/radames/candle-segment-anything-wasm)
 //! - 💻 [GH Link](https://github.com/facebookresearch/segment-anything)
-//! - 📝 [Paper](https://arxiv.org/abs/2304.02643)
+//! - 📝 [Paper](https://huggingface.co/papers/2304.02643)
 //! - 💡 The default backbone can be replaced by the smaller and faster TinyViT model based on [MobileSAM](https://github.com/ChaoningZhang/MobileSAM).
 //!
 //!

candle-transformers/src/models/snac.rs

@@ -4,7 +4,7 @@
 //! See: [SNAC](https://github.com/hubertsiuzdak/snac)
 //!
 /// Multi-Scale Neural Audio Codec (SNAC) compresses audio into discrete codes at a low bitrate.
-/// For more information, read the paper: https://arxiv.org/abs/2410.14411
+/// For more information, read the paper: https://huggingface.co/papers/2410.14411
 ///
 use candle::{DType, Device, IndexOp, Module, Result, Tensor, D};
 use candle_nn::{

candle-transformers/src/models/stable_diffusion/ddim.rs

@@ -6,7 +6,7 @@
 //! this to non-Markovian guidance.
 //!
 //! Denoising Diffusion Implicit Models, J. Song et al, 2020.
-//! https://arxiv.org/abs/2010.02502
+//! https://huggingface.co/papers/2010.02502
 use super::schedulers::{
     betas_for_alpha_bar, BetaSchedule, PredictionType, Scheduler, SchedulerConfig, TimestepSpacing,
 };

candle-transformers/src/models/stable_diffusion/ddpm.rs

@@ -104,7 +104,7 @@ impl DDPMScheduler {
         };
         let current_beta_t = 1. - alpha_prod_t / alpha_prod_t_prev;
 
-        // For t > 0, compute predicted variance βt (see formula (6) and (7) from [the pdf](https://arxiv.org/pdf/2006.11239.pdf))
+        // For t > 0, compute predicted variance βt (see formula (6) and (7) from [the pdf](https://huggingface.co/papers/2006.11239))
         // and sample from it to get previous sample
         // x_{t-1} ~ N(pred_prev_sample, variance) == add variance to pred_sample
         let variance = (1. - alpha_prod_t_prev) / (1. - alpha_prod_t) * current_beta_t;
@@ -112,7 +112,7 @@ impl DDPMScheduler {
         // retrieve variance
         match self.config.variance_type {
             DDPMVarianceType::FixedSmall => variance.max(1e-20),
-            // for rl-diffuser https://arxiv.org/abs/2205.09991
+            // for rl-diffuser https://huggingface.co/papers/2205.09991
             DDPMVarianceType::FixedSmallLog => {
                 let variance = variance.max(1e-20).ln();
                 (variance * 0.5).exp()
@@ -166,12 +166,12 @@ impl DDPMScheduler {
         }
 
         // 4. Compute coefficients for pred_original_sample x_0 and current sample x_t
-        // See formula (7) from https://arxiv.org/pdf/2006.11239.pdf
+        // See formula (7) from https://huggingface.co/papers/2006.11239
         let pred_original_sample_coeff = (alpha_prod_t_prev.sqrt() * current_beta_t) / beta_prod_t;
         let current_sample_coeff = current_alpha_t.sqrt() * beta_prod_t_prev / beta_prod_t;
 
         // 5. Compute predicted previous sample µ_t
-        // See formula (7) from https://arxiv.org/pdf/2006.11239.pdf
+        // See formula (7) from https://huggingface.co/papers/2006.11239
         let pred_prev_sample = ((&pred_original_sample * pred_original_sample_coeff)?
             + sample * current_sample_coeff)?;
 

candle-transformers/src/models/stable_diffusion/resnet.rs

@@ -3,7 +3,7 @@
 //! Some Residual Network blocks used in UNet models.
 //!
 //! Denoising Diffusion Implicit Models, K. He and al, 2015.
-//! - [Paper](https://arxiv.org/abs/1512.03385)
+//! - [Paper](https://huggingface.co/papers/1512.03385)
 //!
 use crate::models::with_tracing::{conv2d, Conv2d};
 use candle::{Result, Tensor, D};

candle-transformers/src/models/stable_diffusion/schedulers.rs

@@ -43,7 +43,7 @@ pub enum PredictionType {
 
 /// Time step spacing for the diffusion process.
 ///
-/// "linspace", "leading", "trailing" corresponds to annotation of Table 2. of the [paper](https://arxiv.org/abs/2305.08891)
+/// "linspace", "leading", "trailing" corresponds to annotation of Table 2. of the [paper](https://huggingface.co/papers/2305.08891)
 #[derive(Debug, Clone, Copy)]
 pub enum TimestepSpacing {
     Leading,

candle-transformers/src/models/stable_diffusion/uni_pc.rs

@@ -10,7 +10,7 @@
 //!
 //! For more information, see the original publication:
 //! UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Models, W. Zhao et al, 2023.
-//! https://arxiv.org/abs/2302.04867
+//! https://huggingface.co/papers/2302.04867
 //!
 //! This work is based largely on UniPC implementation from the diffusers python package:
 //! https://raw.githubusercontent.com/huggingface/diffusers/e8aacda762e311505ba05ae340af23b149e37af3/src/diffusers/schedulers/scheduling_unipc_multistep.py

candle-transformers/src/models/starcoder2.rs

@@ -11,7 +11,7 @@
 //! - Support for 8-bit quantization
 //!
 //! References:
-//! - 📝 [StarCoder Paper](https://arxiv.org/abs/2305.06161)
+//! - 📝 [StarCoder Paper](https://huggingface.co/papers/2305.06161)
 //! - 🤗 [Model Card](https://huggingface.co/bigcode/starcoder)
 //!
 

candle-transformers/src/models/stella_en_v5.rs

@@ -11,7 +11,7 @@
 //! - Rotary positional embeddings (RoPE)
 //!
 //! References:
-//! - [MRL Framework](https://arxiv.org/abs/2205.13147)
+//! - [MRL Framework](https://huggingface.co/papers/2205.13147)
 //! - [Model Card](https://huggingface.co/dunzhang/stella_en_1.5B_v5)
 //!
 
@@ -56,7 +56,7 @@ pub struct Config {
 }
 
 // Excerpt from `stella` model card:
-// `Stella_en_1.5B_v5` models have been trained on [MRL](https://arxiv.org/abs/2205.13147) enabling multiple output dimensions
+// `Stella_en_1.5B_v5` models have been trained on [MRL](https://huggingface.co/papers/2205.13147) enabling multiple output dimensions
 // Embed head represents the config for various embedding dims supported
 #[derive(Debug, Default, Clone, PartialEq, serde::Deserialize)]
 pub struct EmbedHead {

candle-transformers/src/models/t5.rs

@@ -14,7 +14,7 @@
 //! - ⚡ [Interactive Wasm Example](https://huggingface.co/spaces/radames/Candle-T5-Generation-Wasm)
 //! - 💻[GH Model](https://github.com/huggingface/transformers/blob/main/src/transformers/models/t5/modeling_t5.py)
 //! - 🤗 [HF Link](https://huggingface.co/docs/transformers/model_doc/t5)
-//! - 📝 [T5 Paper](https://arxiv.org/abs/1910.10683)
+//! - 📝 [T5 Paper](https://huggingface.co/papers/1910.10683)
 //!
 //! # Encoder-decoder example:
 //!
@@ -33,7 +33,7 @@
 //! # Translation with MADLAD
 //!
 //!
-//! [MADLAD-400](https://arxiv.org/abs/2309.04662) 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.
+//! [MADLAD-400](https://huggingface.co/papers/2309.04662) 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.
 //!
 //! ```bash
 //! cargo run --example t5 --release  -- \

candle-transformers/src/models/trocr.rs

@@ -10,7 +10,7 @@
 //! - Layer normalization and self-attention
 //!
 //! References:
-//! - [Paper](https://arxiv.org/abs/2109.10282)
+//! - [Paper](https://huggingface.co/papers/2109.10282)
 //! - [Model Card](https://huggingface.co/microsoft/trocr-base-handwritten)
 //!
 

candle-transformers/src/models/vgg.rs

@@ -10,7 +10,7 @@
 //! - ReLU activation and dropout
 //!
 //! References:
-//! - [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://arxiv.org/abs/1409.1556)
+//! - [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://huggingface.co/papers/1409.1556)
 //!
 
 use candle::{ModuleT, Result, Tensor};

candle-transformers/src/models/vit.rs

@@ -11,7 +11,7 @@
 //! - Layer normalization
 //!
 //! References:
-//! - [ViT Paper](https://arxiv.org/abs/2010.11929)
+//! - [ViT Paper](https://huggingface.co/papers/2010.11929)
 //! - [Model Card](https://huggingface.co/google/vit-base-patch16-224)
 //!
 

candle-transformers/src/models/yi.rs

@@ -6,7 +6,7 @@
 //! Original code:
 //! - 💻 [Yi Model](https://huggingface.co/01-ai/Yi-6B)
 //! - 💻 [Yi Modeling Code](https://huggingface.co/01-ai/Yi-6B/blob/main/modeling_yi.py)
-//! - 📝 [Technical Report](https://arxiv.org/abs/2403.04652) Yi: Open Foundation Models by 01.AI
+//! - 📝 [Technical Report](https://huggingface.co/papers/2403.04652) Yi: Open Foundation Models by 01.AI
 //!
 //! Key characteristics:
 //! - Multi-head attention with rotary positional embeddings