[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"casestudy-development-moodbench":3},{"id":4,"title":5,"body":6,"client":82,"credits":83,"description":76,"extension":92,"featured":93,"gallery":94,"geo":95,"keyart":97,"lede":99,"mediaType":100,"meta":102,"navigation":103,"path":104,"seo":105,"sort":106,"status":107,"stem":108,"tags":109,"vertical":116,"year":117,"yt_video":118,"__hash__":119},"casestudies\u002Fwork\u002Fdevelopment-moodbench.md","MoodBench",{"type":7,"value":8,"toc":75},"minimark",[9,14,18,21,25,59,63],[10,11,13],"h2",{"id":12},"overview","Overview",[15,16,17],"p",{},"MoodBench is an automated benchmarking framework that fine-tunes, evaluates, and compares small language models for sentiment analysis. It uses Parameter-Efficient Fine-Tuning (PEFT) with LoRA so that meaningful experiments can run on a single Apple Silicon laptop or modest GPU, closing the gap between \"I'd like to evaluate this for my use case\" and \"I have the budget to spin up an H100.\"",[15,19,20],{},"A secondary capability calculates an approximate Net Promoter Score (NPS) over review corpora as a proof-of-concept for downstream business signals layered on top of sentiment classification.",[10,22,24],{"id":23},"whats-interesting-about-it","What's interesting about it",[26,27,28,36,42,48],"ul",{},[29,30,31,35],"li",{},[32,33,34],"em",{},"17 models, one harness."," Coverage spans ultra-tiny (BERT-tiny at 4M params) to medium research-grade (Gemma-2-2B, Pythia-410m), letting you make like-for-like comparisons across architecture families and parameter budgets.",[29,37,38,41],{},[32,39,40],{},"Memory-bounded."," All configurations are designed to fit under 6 GB on Apple Silicon, making the entire suite reproducible without cloud spend.",[29,43,44,47],{},[32,45,46],{},"Thorough metrics."," Beyond accuracy and F1: balanced accuracy, latency percentiles, throughput, memory, statistical-significance testing, and robustness checks.",[29,49,50,53,54,58],{},[32,51,52],{},"Two interfaces."," A CLI (",[55,56,57],"code",{},"uv run moodbench …",") for CI\u002FCD and reproducible runs, and a Gradio web UI for interactive exploration.",[10,60,62],{"id":61},"references","References",[26,64,65],{},[29,66,67,68],{},"Repository: ",[69,70,74],"a",{"href":71,"rel":72},"https:\u002F\u002Fgithub.com\u002Fandrewmarconi\u002FMoodBench",[73],"nofollow","github.com\u002Fandrewmarconi\u002FMoodBench",{"title":76,"searchDepth":77,"depth":77,"links":78},"",2,[79,80,81],{"id":12,"depth":77,"text":13},{"id":23,"depth":77,"text":24},{"id":61,"depth":77,"text":62},"Personal Project",[84,89],{"sort":85,"role":86,"value":87,"uri":88},1,"Creator & Developer","Andrew Marconi","https:\u002F\u002Flinkedin.com\u002Fin\u002Fmarconi",{"sort":90,"role":91,"value":74,"uri":71},10,"Repository","md",false,[],[96],"Global",{"src":98,"alt":5},"\u002Fwork\u002Fdevelopment-moodbench\u002Fmoodbench-keyart.jpg","A multi-LLM sentiment-analysis benchmark: fine-tunes, evaluates, and compares 17 small language models (4M to 410M parameters) on consumer hardware using LoRA.",[101],"CLI Tool",{},true,"\u002Fwork\u002Fdevelopment-moodbench",{"title":5,"description":76},190,"live","work\u002Fdevelopment-moodbench",[110,111,112,113,114,115],"Python","LoRA","PEFT","NLP","Benchmarking","Open Source","AI\u002FMachine Learning",2025,null,"vG5iv_AlQVJT462bwao8HJMnLs7RvcxHXnephYv2GWc"]