hharsha/agentic-github-tagger

1---2language:3  - en4license: apache-2.05library_name: transformers6base_model: google-t5/t5-small7tags:8  - peft9  - t510  - agents11  - rag12  - llmops13  - lora14  - text2text-generation15datasets:16  - hharsha/agentic-github-meta17pipeline_tag: text2text-generation18widget:19  - text: multi-agent platform with RAG, MCP, and observability20  - text: looking for RAG hybrid search recall@k and reranking21  - text: FastAPI Next.js agent backend with Docker Compose22---23 24# agentic-github-tagger25 26Lightweight **text2text tag generator** for agentic AI / RAG / LLMOps GitHub-style27descriptions. Fine-tuned from [`google-t5/t5-small`](https://huggingface.co/google-t5/t5-small)28with **PEFT LoRA** (r=16, alpha=32, dropout=0.05, target_modules `q`,`v`) on29[`hharsha/agentic-github-meta`](https://huggingface.co/datasets/hharsha/agentic-github-meta),30then **merged** so full small weights load on free CPU.31 32> ~60M-param T5-small tagger — **not** a 7B chat demo. Free Hub + CPU friendly.33 34## Usage35 36```python37from transformers import AutoModelForSeq2SeqLM, AutoTokenizer38 39model_id = "hharsha/agentic-github-tagger"40tok = AutoTokenizer.from_pretrained(model_id)41model = AutoModelForSeq2SeqLM.from_pretrained(model_id)42 43text = "multi-agent platform with RAG, MCP, and observability"44ids = tok(text, return_tensors="pt")45out = model.generate(**ids, max_new_tokens=64, num_beams=4)46print(tok.decode(out[0], skip_special_tokens=True))47```48 49On older `transformers` that still register the task, this also works:50 51```python52from transformers import pipeline53pipe = pipeline("text2text-generation", model="hharsha/agentic-github-tagger")54print(pipe("multi-agent platform with RAG, MCP, and observability")[0]["generated_text"])55```56 57Sample output from this training run:58 59```60multi-agent, multi-agent, observability, rag, mCP, observability61```62 63## Training64 65| | |66|---|---|67| Base | `google-t5/t5-small` |68| Method | PEFT LoRA then merge |69| r / alpha / dropout | 16 / 32 / 0.05 |70| target_modules | q, v |71| Epochs | 3 (CPU) |72| Batch size | 8 |73| Dataset | [`hharsha/agentic-github-meta`](https://huggingface.co/datasets/hharsha/agentic-github-meta) (687 rows; 600 used for train) |74 75## Links76 77- Dataset: [`hharsha/agentic-github-meta`](https://huggingface.co/datasets/hharsha/agentic-github-meta)78- Showcase: [`hharsha/agentic-systems-showcase`](https://huggingface.co/datasets/hharsha/agentic-systems-showcase)79- Studio: [https://agentic-systems-studio.com](https://agentic-systems-studio.com)80- GitHub: [https://github.com/hharsha98](https://github.com/hharsha98)81 82## Intended use / limits83 84Auto-suggest comma-separated tags for agentic / RAG / LLMOps project listings.85Small model; tags can repeat or be incomplete. Not for safety-critical labeling.86