Pinned Loading

1. sign-language-translation sign-language-translation Public

   video-to-text sign language translation with a modified SAM-3D-Body library fork, evaluated on How2Sign

   Jupyter Notebook

2. imagenet-classification-mobilint-mla-100 imagenet-classification-mobilint-mla-100 Public

   ImageNet (subset) Classification on Mobilint's MLA-100 Edge NPU

   Jupyter Notebook

3. manifests manifests Public

   Kubernetes manifest files to easily self host at scale with ArgoCD. Includes configs for n8n, langflow, langfuse, loki + otel, mcp forge, getOutline, argo workflows, superset and more

4. bert-tweeteval bert-tweeteval Public

   BERT is an encoder-transformer based architecture that is more suitable for some tasks, such as classification. We benchmark both base and fine-tuned performance trained on the TweetEval dataset.

   Jupyter Notebook

5. wixqa-rag wixqa-rag Public

   Building and evaluating Retrieval-Augmented Generation system for the WixQA dataset with Cerebras and gpt-oss-120b. We evaluate different chunking and retrieval strategies k={1,3,5,10}, chunks={300…

   Jupyter Notebook

6. mamba mamba Public

   Forked from state-spaces/mamba

   Mamba SSM architecture (forked for AGX Orin compilation). With 130M params, we achieve 225 tokens per second, which is constant regardless of context length, due to space-state architecture.

   Python