Spaces:
Running
Running
README: add model card highlights section and metadata snapshot.
Browse files
README.md
CHANGED
|
@@ -30,6 +30,19 @@ Deterministic OpenEnv benchmark for real SQL debugging workflows. This project e
|
|
| 30 |
- GitHub: [https://github.com/mdayan8/sql-debug-env](https://github.com/mdayan8/sql-debug-env)
|
| 31 |
- W&B dashboard: [https://wandb.ai/mdayanbag-pesitm/sql-debug-grpo-best-budget/workspace?nw=nwusermdayanbag](https://wandb.ai/mdayanbag-pesitm/sql-debug-grpo-best-budget/workspace?nw=nwusermdayanbag)
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
## Problem and Motivation
|
| 34 |
|
| 35 |
SQL debugging is expensive, repetitive, and operationally risky:
|
|
|
|
| 30 |
- GitHub: [https://github.com/mdayan8/sql-debug-env](https://github.com/mdayan8/sql-debug-env)
|
| 31 |
- W&B dashboard: [https://wandb.ai/mdayanbag-pesitm/sql-debug-grpo-best-budget/workspace?nw=nwusermdayanbag](https://wandb.ai/mdayanbag-pesitm/sql-debug-grpo-best-budget/workspace?nw=nwusermdayanbag)
|
| 32 |
|
| 33 |
+
## Model Card Highlights
|
| 34 |
+
|
| 35 |
+
Model: [md896/sql-debug-agent-qwen25-05b-grpo-wandb-continue-v2](https://huggingface.co/md896/sql-debug-agent-qwen25-05b-grpo-wandb-continue-v2)
|
| 36 |
+
|
| 37 |
+
| Field | Value |
|
| 38 |
+
|---|---|
|
| 39 |
+
| Task | Text generation (SQL repair style prompts) |
|
| 40 |
+
| Libraries | Transformers, TRL (GRPO), Safetensors, TGI-compatible |
|
| 41 |
+
| Family tags | qwen2, grpo, conversational, text-generation-inference |
|
| 42 |
+
| Base tracks used in workflow | Qwen2.5-Coder 0.5B bridge + Qwen2.5-Coder 7B benchmark/eval track |
|
| 43 |
+
| Training signal | Execution-grounded reward from OpenEnv SQL tasks |
|
| 44 |
+
| Reference | arXiv:1910.09700 (as listed in model metadata) |
|
| 45 |
+
|
| 46 |
## Problem and Motivation
|
| 47 |
|
| 48 |
SQL debugging is expensive, repetitive, and operationally risky:
|