Merge branch 'main' into trackio

Author: abidlabs

_blog.yml

@@ -6340,6 +6340,7 @@
     - models
     - open-source
 
+
 - local: migrating-the-hub-to-xet
   title: "Migrating the Hub from Git LFS to Xet"
   author: jsulz
@@ -6397,6 +6398,7 @@
 - local: consilium-multi-llm
   title: "Consilium: When Multiple LLMs Collaborate"
   author: azettl
+  guest: true
   thumbnail: /blog/assets/consilium-multi-llm/thumbnail.png
   date: Jul 17, 2025
   tags:
@@ -6413,3 +6415,49 @@
   tags:
     - collaboration
     - guide
+
+- local: lora-fast 
+  title: "Fast LoRA inference for Flux with Diffusers and PEFT"
+  thumbnail: /blog/assets/lora-fast/thumbnail.png
+  author: sayakpaul
+  date: July 23, 2025
+  tags:
+    - lora
+    - diffusion
+    - guide
+
+- local: timescope-video-lmm-benchmark
+  title: "TimeScope: How Long Can Your Video Large Multimodal Model Go?"
+  author: orrzohar
+  thumbnail: /blog/assets/timescope/thumbnail.png
+  date: Jul 23, 2025
+  tags:
+    - video
+    - datasets
+    - multimodal
+    - open-source
+    - benchmark
+    
+- local: parquet-cdc
+  title: "Parquet Content-Defined Chunking"
+  author: kszucs
+  thumbnail: /blog/assets/parquet-cdc/thumbnail.png
+  date: July 25, 2025
+  tags:
+    - data
+    - datasets
+    - dedupe
+    - hub
+    - parquet
+    - storage
+    - xet
+
+- local: hf-cli
+  title: "Say hello to `hf`: a faster, friendlier Hugging Face CLI ✨"
+  author: Wauplin
+  thumbnail: /blog/assets/hf-cli-thumbnail.png
+  date: Jul 25, 2025
+  tags:
+    - cli
+    - huggingface_hub
+    - python

assets/hf-cli-thumbnail.png

@@ -3,6 +3,7 @@ title: "Consilium: When Multiple LLMs Collaborate"
 thumbnail: /blog/assets/consilium-multi-llm/thumbnail.png
 authors:
 - user: azettl
+  guest: true
 ---
 
 # Consilium: When Multiple LLMs Collaborate

assets/lora-fast/thumbnail.png

@@ -406,7 +406,7 @@ of values from our input vector. For \\(2D\\) data, we need to encode both horiz
 
 ## The future of positional encoding
 
-Is RoPE the final incarnation of positional encoding? This [recent paper](https://arxiv.org/pdf/2410.06205) from DeepMind deeply analyses RoPE and highlights some fundamental problems. TLDR: RoPE isn't a perfect solution, and the models mostly focus on the lower frequencies and the rotation for a certain percent of low frequencies improves performance on Gemma 2B! 
+Is RoPE the final incarnation of positional encoding? This [recent paper](https://arxiv.org/pdf/2410.06205) from DeepMind deeply analyses RoPE and highlights some fundamental problems. TLDR: RoPE isn't a perfect solution, and the models mostly focus on the lower frequencies, but the paper shows that **removing** (not rotating) the lowest frequencies improves performance on Gemma 2B! 
 
 I anticipate some future breakthroughs, perhaps taking inspiration from
 signal processing with ideas like wavelets or hierarchical implementations. As models

assets/parquet-cdc/thumbnail.png

@@ -807,4 +807,4 @@ The complete Ettin suite includes models at six different scales (for both encod
 - [📝 Paper](https://github.com/jhu-clsp/ettin-encoder-vs-decoder)  
 - [🗂️ Training Data](https://huggingface.co/datasets/jhu-clsp/ettin-pretraining-data) (2T+ tokens, fully open)
 - [💻 GitHub Repository](https://github.com/jhu-clsp/ettin-encoder-vs-decoder)
-- [📊 250+ Training Checkpoints](https://huggingface.co/datasets/jhu-clsp/ettin-checkpoints) for studying training dynamics or knowledge learning
+- [📊 250+ Training Checkpoints](https://huggingface.co/jhu-clsp/ettin-checkpoints) for studying training dynamics or knowledge learning

assets/timescope/thumbnail.png

@@ -40,7 +40,7 @@ Forecasting makes **contamination impossible by design**, as you can't train on
 
 Perhaps most importantly, predictions about the future are **inherently verifiable**. We can wait and see who was right, creating an objective, time-stamped measure of model performance.
 
-We therefore propose evaluating agents on their ability to predict future events (Ye et al., 2024; Karger et al., 2025). **FutureBench** draws from real-world prediction markets and emerging news to create interesting prediction tasks grounded in actual future outcomes. We collect events from platforms and live news coverages and manifolds markets, filtering them to focus on emerging events worth predicting. Using an agent-based approach, we curate scenarios that require genuine reasoning rather than simple pattern matching. Think geopolitical developments, market movements, or technology adoption trends - events where informed analysis actually matters.
+We therefore propose evaluating agents on their ability to predict future events (Ye et al., 2024; Karger et al., 2025). **FutureBench** draws from real-world prediction markets and emerging news to create interesting prediction tasks grounded in actual future outcomes. We collect events from platforms and live news coverages and manifold markets, filtering them to focus on emerging events worth predicting. Using an agent-based approach, we curate scenarios that require genuine reasoning rather than simple pattern matching. Think geopolitical developments, market movements, or technology adoption trends - events where informed analysis actually matters.
 
 ## Can Agents Predict Future Events?
 This is the obvious question, and it's at the heart of what makes this benchmark interesting! We believe the answer cannot be a simple “yes” or a “no”, as it mostly depends on the actual questions; there are always important caveats to consider.

consilium-multi-llm.md

@@ -37,7 +37,7 @@ As an MCP developer, it's highly recommended to implement your MCP tools to emit
 
 ## Transform OpenAPI Specs to MCP in One Line
 
-If you want to integrate an existing backend API into an LLM, you have to manually map API endpoints to MCP tools. This can be a time consuming and error prone chore. With this release, Gradio can automate the entire process for you! With a single line of code, you can integrate your business backend into any MCP-compatible LLM. 
+If you want to integrate an existing backend API into an LLM, you have to manually map API endpoints to MCP tools. This can be a time-consuming and error prone chore. With this release, Gradio can automate the entire process for you! With a single line of code, you can integrate your business backend into any MCP-compatible LLM. 
 
 [OpenAPI](https://www.openapis.org/) is a widely adopted standard for describing RESTful APIs in a machine-readable format, typically as a JSON file. Gradio now features the `gr.load_openapi` function, which creates a Gradio application directly from an OpenAPI schema. You can then launch the app with `mcp_server=True` to automatically create an MCP server for your API!
 
@@ -96,7 +96,7 @@ You can read more about this in the Gradio [Guides](https://www.gradio.app/guide
 
 ## Modifying Tool Descriptions
 
-Gradio automatically generates tool descriptions from your function names and docstrings. Now you can customize the tool description even further with the `api_description` parmeter. In this example, the tool description will read "Apply a sepia filter to any image."
+Gradio automatically generates tool descriptions from your function names and docstrings. Now you can customize the tool description even further with the `api_description` parameter. In this example, the tool description will read "Apply a sepia filter to any image."
 
 ```python
 import gradio as gr
@@ -129,4 +129,4 @@ Read more in the [guide](https://www.gradio.app/guides/building-mcp-server-with-
 
 ## Conclusion
 
-Want us to add a new MCP-related feature to Gradio? Let us know in the comments of the blog or on [GitHub](https://github.com/gradio-app/gradio/issues). Also if you've built a cool MCP server or Gradio app let us know in the comments and we'll amplify it!
+Want us to add a new MCP-related feature to Gradio? Let us know in the comments of the blog or on [GitHub](https://github.com/gradio-app/gradio/issues). Also if you've built a cool MCP server or Gradio app let us know in the comments and we'll amplify it!