Merge remote-tracking branch 'upstream/main'

Author: thomas-infisical

_blog.yml

@@ -5426,7 +5426,7 @@
   tags:
     - aws
     - partnerships
-  
+
 - local: ai-art-newsletter-jan-25
   title: "The AI tools for Art Newsletter - Issue 1"
   author: linoyts
@@ -5439,7 +5439,7 @@
     - community
 
 - local: dabstep
-  title: "DABStep: Data Agent Benchmark for Multi-step Reasoning" 
+  title: "DABStep: Data Agent Benchmark for Multi-step Reasoning"
   thumbnail: /blog/assets/dabstep/thumbnail.png
   author: eggie5
   guest: True
@@ -5450,7 +5450,6 @@
     - research
     - evaluation
 
-
 - local: pi0
   title: "π0 and π0-FAST: Vision-Language-Action Models for General Robot Control"
   author: danaaubakirova
@@ -5462,7 +5461,7 @@
     - community
 
 - local: open-deep-research
-  title: "Open-source DeepResearch – Freeing our search agents" 
+  title: "Open-source DeepResearch – Freeing our search agents"
   thumbnail: /blog/assets/open-deep-research/thumbnail.png
   author: m-ric
   date: Feb 4, 2025
@@ -5472,19 +5471,6 @@
     - research
     - smolagents
 
-- local: scaling-secrets-management
-  title: "How Hugging Face Scaled Secrets Management for AI Infrastructure"  
-  thumbnail: /blog/assets/infisical/thumbnail.png
-  author: segudev
-  guest: true
-  date: Feb 10, 2025
-  tags:
-    - secrets
-    - security
-    - shift-left
-    - infrastructure
-    - open-source
-    
 - local: leaderboard-arabic-v2
   title: "The Open Arabic LLM Leaderboard 2"
   thumbnail: /blog/assets/leaderboards-on-the-hub/thumbnail_arabic.png
@@ -5497,4 +5483,116 @@
     - leaderboard
     - LLM
     - arabic
-    
+
+- local: vid_ds_scripts
+  title: "Build awesome datasets for video generation"
+  author: hlky
+  thumbnail: /blog/assets/vid_ds_scripts/thumbnail.png
+  date: Feb 12, 2025
+  tags:
+    - guide
+    - video
+    - datasets
+
+- local: from-chunks-to-blocks
+  title: "From Chunks to Blocks: Accelerating Uploads and Downloads on the Hub"
+  author: jsulz
+  thumbnail: /blog/assets/from-chunks-to-blocks/thumbnail.png
+  date: February 12, 2025
+  tags:
+    - dedupe
+    - storage
+    - content defined chunking
+    - quantization
+
+- local: billion-classifications
+  title: "1 Billion Classifications"
+  author: derek-thomas
+  thumbnail: /blog/assets/billion-classifications/billion-classifications-thumbnail.png
+  guest: true
+  date: Feb 13, 2025
+  tags:
+    - inference-endpoints
+    - classification
+    - embedding
+    - embeddings
+    - nlp
+    - python
+    - cost
+    - enterprise
+
+- local: math_verify_leaderboard
+  title: "Fixing Open LLM Leaderboard with Math-Verify"
+  author: hynky
+  thumbnail: /blog/assets/math_verify_leaderboard/thumbnail.png
+  date: Feb 14, 2025
+  tags:
+    - math-verify
+    - open-llm-leaderboard
+    - leaderboard
+    - evaluation
+
+- local: fireworks-ai
+  title: "Welcome Fireworks.ai on the Hub 🎆"
+  author: julien-c
+  thumbnail: /blog/assets/inference-providers/welcome-fireworks.jpg
+  date: Feb 14, 2025
+  tags:
+    - announcement
+    - hub
+
+- local: inference-providers-nebius-novita-hyperbolic
+  title: "Introducing Three New Serverless Inference Providers: Hyperbolic, Nebius AI Studio, and Novita 🔥"
+  author: reach-vb
+  thumbnail: /blog/assets/inference-providers/second-batch-thumbnail.webp
+  date: Feb 18, 2025
+  tags:
+    - announcement
+    - hub
+
+- local: paligemma2mix
+  title: "PaliGemma 2 Mix - New Instruction Vision Language Models by Google"
+  thumbnail: /blog/assets/paligemma2/thumbnail.png
+  author: ariG23498
+  date: Feb 19, 2025
+  tags:
+    - multimodal
+    - LLM
+    - vision
+
+- local: smolvlm2
+  title: "SmolVLM2: Bringing Video Understanding to Every Device"
+  author: orrzohar
+  guest: true
+  thumbnail: /blog/assets/smolvlm2/banner.png
+  date: Feb 20, 2025
+  tags:
+    - vlm
+    - multimodal
+    - video
+    - on-device
+    - llm
+    - nlp
+    - vision
+
+- local: siglip2
+  title: "SigLIP 2: A better multilingual vision language encoder"
+  author: ariG23498
+  thumbnail: /blog/assets/siglip2/thumbnail.png
+  date: Feb 21, 2025
+  tags:
+    - multimodal
+    - vision
+
+- local: scaling-secrets-management
+  title: "How Hugging Face Scaled Secrets Management for AI Infrastructure"  
+  thumbnail: /blog/assets/infisical/thumbnail.png
+  author: segudev
+  guest: true
+  date: Feb 21, 2025
+  tags:
+    - secrets
+    - security
+    - shift-left
+    - infrastructure
+    - open-source    

assets/billion-classifications/billion-classifications-thumbnail.png

@@ -8,7 +8,7 @@ authors:
   guest: True
 - user: frisokingma
   guest: True
-- user: andreu-adyen
+- user: andreumora
   guest: True
 - user: lvwerra
 - user: thomwolf

assets/from-chunks-to-blocks/thumbnail.png

@@ -24,6 +24,7 @@ We collaborate with Amazon Web Services to make it easier for developers to depl
 Let’s review how you can deploy and fine-tune DeepSeek R1 models with Hugging Face on AWS.
 - [Deploy DeepSeek R1 models](#deploy-deepseek-r1-models)
     - [Deploy on AWS with Hugging Face Inference Endpoints](#deploy-on-aws-with-hugging-face-inference-endpoints)
+    - [Deploy on Amazon Bedrock Marketplace]
     - [Deploy on Amazon SageMaker AI with Hugging Face LLM DLCs](#deploy-on-amazon-sagemaker-ai-with-hugging-face-llm-dlcs)
         - [DeepSeek R1 on GPUs](#deepseek-r1-on-gpus)
         - [Distilled models on GPUs](#distilled-models-on-gpus)
@@ -48,6 +49,12 @@ You can find DeepSeek R1 and distilled models, as well as other popular open LLM
 
 | **Note:** The team is working on enabling DeepSeek models deployment on Inferentia instances. Stay tuned!
 
+### Deploy on Amazon Bedrock Marketplace
+
+You can deploy the Deepseek distilled models on Amazon Bedrock via the marketplace, which will deploy an endpoint in Amazon SageMaker AI under the hood. Here is a video of how you can navigate through the AWS console:
+
+![bedrock-deployment.gif](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/deepseek-aws/bedrock-deployment.gif)
+
 ### Deploy on Amazon Sagemaker AI with Hugging Face LLM DLCs
 
 #### DeepSeek R1 on GPUs
@@ -56,7 +63,12 @@ You can find DeepSeek R1 and distilled models, as well as other popular open LLM
 
 #### Distilled models on GPUs
 
-Let’s walk through the deployment of DeepSeek-R1-Distill-Llama-70B. 
+You can deploy the Deepseek distilled models on Amazon Sagemaker AI with Hugging Face LLM DLCs using Jumpstart directly or using the Python Sagemaker SDK.
+Here is a video of how you can navigate through the AWS console:
+
+![jumpstart-deployment.gif](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/blog/deepseek-aws/jumpstart-deployment.gif)
+
+Now we have seen how to deploy usig Jumpstart, let’s walk through the Python Sagemaker SDK deployment of DeepSeek-R1-Distill-Llama-70B. 
 
 Code snippets are available on the model page under the Deploy button! 
 

assets/inference-providers/second-batch-thumbnail.webp

@@ -0,0 +1,129 @@
+---
+title: "Welcome Fireworks.ai on the Hub 🎆"
+thumbnail: /blog/assets/inference-providers/welcome-fireworks.jpg
+authors:
+- user: teofeliu
+  guest: true
+  org: fireworks-ai
+- user: shaunak-fireworks
+  guest: true
+  org: fireworks-ai
+- user: julien-c
+---
+
+Following our recent announcement on [Inference Providers on the Hub](https://huggingface.co/blog/inference-providers), we're thrilled to share that **Fireworks.ai** is now a supported Inference Provider on HF Hub!
+
+[Fireworks.ai](https://fireworks.ai) delivers blazing-fast serverless inference directly on model pages, as well as throughout the whole HF ecosystem of libraries and tools, making it easier than ever to run inference on your favorite models.
+
+<img src="https://huggingface.co/blog/assets/inference-providers/welcome-fireworks.jpg" alt="Fireworks.ai supported as Inference Provider on Hugging Face"/>
+
+Among others, starting now, you can run serverless inference to the following models via Fireworks.ai:
+
+- [deepseek-ai/DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1)
+- [deepseek-ai/DeepSeek-V3](https://huggingface.co/deepseek-ai/DeepSeek-V3)
+- [mistralai/Mistral-Small-24B-Instruct-2501](https://huggingface.co/mistralai/Mistral-Small-24B-Instruct-2501)
+- [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)
+- [meta-llama/Llama-3.2-90B-Vision-Instruct](https://huggingface.co/meta-llama/Llama-3.2-90B-Vision-Instruct)
+
+and many more, you can find the full list [here](https://huggingface.co/models?inference_provider=fireworks-ai).
+
+Light up your projects with Fireworks.ai today!
+
+## How it works
+
+### In the website UI
+
+![Fireworks.ai inference provider UI](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/inference-providers/fireworks.png)
+
+Search for all models supported by Fireworks on HF **[here](https://huggingface.co/models?inference_provider=fireworks-ai)**.
+
+### From the client SDKs
+
+#### from Python, using huggingface_hub
+
+The following example shows how to use DeepSeek-R1 using Fireworks.ai as your inference provider. You can use a [Hugging Face token](https://huggingface.co/settings/tokens) for automatic routing through Hugging Face, or your own Fireworks.ai API key if you have one.
+
+Install `huggingface_hub` from source: 
+
+```bash
+pip install git+https://github.com/huggingface/huggingface_hub
+```
+
+Use the `huggingface_hub` python library to call Fireworks.ai endpoints by defining the `provider` parameter.
+
+```python
+from huggingface_hub import InferenceClient
+
+client = InferenceClient(
+    provider="fireworks-ai",
+    api_key="xxxxxxxxxxxxxxxxxxxxxxxx"
+)
+
+messages = [
+    {
+        "role": "user",
+        "content": "What is the capital of France?"
+    }
+]
+
+completion = client.chat.completions.create(
+    model="deepseek-ai/DeepSeek-R1", 
+    messages=messages, 
+    max_tokens=500
+)
+
+print(completion.choices[0].message)
+```
+
+#### from JS using @huggingface/inference
+
+```js
+import { HfInference } from "@huggingface/inference";
+
+const client = new HfInference("xxxxxxxxxxxxxxxxxxxxxxxx");
+
+const chatCompletion = await client.chatCompletion({
+    model: "deepseek-ai/DeepSeek-R1",
+    messages: [
+        {
+            role: "user",
+            content: "How to make extremely spicy Mayonnaise?"
+        }
+    ],
+    provider: "fireworks-ai",
+    max_tokens: 500
+});
+
+console.log(chatCompletion.choices[0].message);
+```
+
+### From HTTP calls
+
+Here's how you can call Llama-3.3-70B-Instruct using Fireworks.ai as the inference provider via cURL.
+
+```
+curl 'https://router.huggingface.co/fireworks-ai/v1/chat/completions' \
+-H 'Authorization: Bearer xxxxxxxxxxxxxxxxxxxxxxxx' \
+-H 'Content-Type: application/json' \
+--data '{
+    "model": "accounts/fireworks/models/llama-v3p3-70b-instruct",
+    "messages": [
+        {
+            "role": "user",
+            "content": "What is the meaning of life if you were a dog?"
+        }
+    ],
+    "max_tokens": 500,
+    "stream": false
+}'
+```
+
+## Billing
+
+For direct requests, i.e. when you use a Fireworks key, you are billed directly on your Fireworks account.
+
+For routed requests, i.e. when you authenticate via the hub, you'll only pay the standard Fireworks API rates. There's no additional markup from us, we just pass through the provider costs directly. (In the future, we may establish revenue-sharing agreements with our provider partners.)
+
+Important Note ‼️ PRO users get $2 worth of Inference credits every month. You can use them across providers. 🔥
+
+Subscribe to the [Hugging Face PRO plan](https://hf.co/subscribe/pro) to get access to Inference credits, ZeroGPU, Spaces Dev Mode, 20x higher limits, and more.