|
Download README.md from VanzCoder/Final_Assignment_Template1: direct link, hf CLI and curl.
- Browser
- Download file 1.97 kB
-
https://huggingface.co/spaces/VanzCoder/Final_Assignment_Template1/resolve/main/README.md
- Command line
-
hf download hf://spaces/VanzCoder/Final_Assignment_Template1/README.md
-
curl -L -o README.md https://huggingface.co/spaces/VanzCoder/Final_Assignment_Template1/resolve/main/README.md
1.97 kB
| title: Template Final Assignment | |
| emoji: π΅π»ββοΈ | |
| colorFrom: indigo | |
| colorTo: indigo | |
| sdk: gradio | |
| sdk_version: 5.25.2 | |
| app_file: app.py | |
| pinned: false | |
| hf_oauth: true | |
| hf_oauth_expiration_minutes: 480 | |
| # GAIA Agent β Final Assignment | |
| A `smolagents` CodeAgent built to answer GAIA Level 1 questions for the HF | |
| Agents course final assignment. Uses HF Inference Providers for the LLM and a | |
| small toolset for browsing, files, audio, and images. Self-consistency + | |
| verifier pass on top to maximize exact-match score. | |
| ## Where things live | |
| | File | Purpose | | |
| | ------------- | -------------------------------------------------------- | | |
| | `app.py` | Gradio UI + run-and-submit pipeline | | |
| | `agent.py` | `GaiaAgent` class β self-consistency + verifier | | |
| | `tools.py` | 8 tools (search, browse, wiki, YT, file, table, ASR, VLM)| | |
| | `config.py` | **All non-secret settings** (model IDs, knobs) | | |
| | `requirements.txt` | Python deps | | |
| ## Secrets (Space β Settings β Variables and secrets) | |
| | Secret | Purpose | | |
| | ---------------- | -------------------------------------------------------- | | |
| | `HF_TOKEN` | HF Inference Providers (uses your Pro quota) | | |
| | `SERPER_API_KEY` | Web search via serper.dev | | |
| That's it. No model IDs in secrets β edit `config.py` for those. | |
| ## Tuning | |
| Open `config.py` and change values directly: | |
| - `AGENT_MODEL_ID` β the LLM brain. Default DeepSeek-V3. | |
| - `SELF_CONSISTENCY_N` β independent attempts per question. 3 = balanced, 5 = max. | |
| - `MAX_STEPS` β agent reasoning steps per attempt. | |
| - `VLM_MODEL_ID` β vision model for image questions. | |
| Any of the constants in `config.py` can also be overridden by a Space env var | |
| of the same name, but that's optional. | |