Alan Ponnachan

Errors with Attention Mechanisms in Janus Inference Script (FlashAttention 2.0 & _flash_supports_window_size)

3

I encountered multiple issues when running the Janus-1.3B model [inference script](https://github.com/deepseek-ai/Janus/blob/main/inference.py), both with and without enabling FlashAttention. These errors prevent successful execution of the model in a standard environment, such...

Add Notebook: Documentation Chatbot with Meta Synthetic Data Kit

1

# What does this PR do? Fixes https://github.com/huggingface/cookbook/issues/293 This pull request adds a new notebook to the Hugging Face Cookbook: fine_tune_chatbot_docs_synthetic.ipynb. Provides a hands-on example on [Meta's synthetic-data-kit](https://github.com/meta-llama/synthetic-data-kit) ## Who...

I propose adding a new notebook to the Hugging Face Cookbook that demonstrates the [Meta's synthetic-data-kit](https://github.com/meta-llama/synthetic-data-kit) to automatically generate Question/Answer pairs and further use [Unsloth](https://github.com/unslothai/unsloth) for finetuning model on generated...

fix(bedrock): Add prompt caching support for Converse API

10

- [x] I have added tests that cover my changes. - [ ] If adding a new instrumentation or changing an existing one, I've added screenshots from some observability platform...

fix(vertex_ai): Resolve JSONDecodeError in Gemini streaming

2

## Resolve JSONDecodeError in Gemini streaming ## Relevant issues Fixes #16562 ## Pre-Submission checklist **Please complete all items before asking a LiteLLM maintainer to review your PR** - [x] I...

feat: Add Adversarial Robustness Metric

2

Fixes #2150 #### Description This PR introduces the `AdversarialRobustnessMetric`, a new metric designed to evaluate the resilience of Large Language Models (LLMs) against common textual perturbations. This directly addresses the...

**Is your feature request related to a problem? Please describe.** LLMs are becoming central to many applications, but their reliability is often undermined by their vulnerability to adversarial perturbations. Small,...

feat(metrics): Add Risk-Control Metric Suite

3

## Issue Link / Problem Description - Fixes #2279 - **Problem:** The `ragas` library currently excels at evaluating the quality of generated answers but lacks metrics to assess a RAG...

feat: Addition of Risk-Control Metrics for Trustworthy RAG Evaluation

1

### Describe the Feature This proposal introduces a new, cohesive suite of four interconnected and academically-grounded metrics designed to evaluate a critical dimension currently missing from `ragas`: **a RAG system's...

## Allow dynamic configuration of built-in tools via RunContext This PR closes #3555 Adding support for dynamic configuration of built-in tools (e.g., `WebSearchTool`) based on the agent's `RunContext`. Previously, `builtin_tools`...