ZhSQL-R1-1.5B is an open-weight model from Rain. Its published files total 3.9 GB.
ZhSQL-R1-1.5B is a bilingual Text-to-SQL model based on Qwen2.5-Coder-1.5B, trained using supervised fine-tuning (SFT) followed by execution-reward GRPO.
Model Card
ZhSQL-R1-1.5B is a bilingual Text-to-SQL model based on Qwen2.5-Coder-1.5B, trained using supervised fine-tuning (SFT) followed by execution-reward GRPO. The model was fine-tuned for two epochs on 7,127 examples derived from the BIRD benchmark, with approximately 48% Chinese prompts. Key settings: LoRA rank 32, maximum sequence length 3,072, effective batch size 16, and learning rate 2e-4. Loss was applied only to the SQL response. The SFT model generated eight SQL candidates per question. Examples with partial success (0 < passrate < 1) were retained, producing 3,692 learnable examples. GRPO used binary execution-equivalence rewards for 300 steps, with eight generations per prompt, LoRA…
Excerpt from the card by Rain.
Identity and Version
- Repository
- Rainyyyyy/ZhSQL-R1-1.5B
- Publisher
- Rain
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- en, zh
- Revision
- 4bf9ed0fef3626b7c5fa33293b7862c730d11e18
- First published
- 2026-09-21
- Last updated
- 2026-09-21
Files and Weights
48 files, 3.9 GB in total. The weights are 17 files totalling 3.8 GB in bin, pt, pth, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| lora/adapter_model.safetensors | Weights | 73.9 MB | b2cec12121b5 |
| merged/model.safetensors | Weights | 3.1 GB | c00213f46478 |
| trainer/checkpoint-100/adapter_model.safetensors | Weights | 147.8 MB | 01cc9ab62c09 |
| trainer/checkpoint-100/optimizer.pt | Weights | 75.5 MB | 0df7a4421d41 |
| trainer/checkpoint-100/rng_state.pth | Weights | 14.6 KB | 081d189107a3 |
| trainer/checkpoint-100/scheduler.pt | Weights | 1.5 KB | 90050df4c702 |
| trainer/checkpoint-100/training_args.bin | Weights | 6.6 KB | 2cd6298d0055 |
| trainer/checkpoint-200/adapter_model.safetensors | Weights | 147.8 MB | f1fa85e4d95f |
| trainer/checkpoint-200/optimizer.pt | Weights | 75.5 MB | 7e5640a93621 |
| trainer/checkpoint-200/rng_state.pth | Weights | 14.6 KB | f386eb617c8e |
| trainer/checkpoint-200/scheduler.pt | Weights | 1.5 KB | a4ecd5301e10 |
| trainer/checkpoint-200/training_args.bin | Weights | 6.6 KB | 2cd6298d0055 |
| trainer/checkpoint-300/adapter_model.safetensors | Weights | 147.8 MB | 88032b6c1ff5 |
| trainer/checkpoint-300/optimizer.pt | Weights | 75.5 MB | 6734a09f2553 |
| trainer/checkpoint-300/rng_state.pth | Weights | 14.6 KB | 690e83f5a9fb |
| trainer/checkpoint-300/scheduler.pt | Weights | 1.5 KB | 73803a757f07 |
| trainer/checkpoint-300/training_args.bin | Weights | 6.6 KB | 2cd6298d0055 |
| log_history.json | Configuration | 271.5 KB | — |
| lora/adapter_config.json | Configuration | 1.3 KB | — |
| merged/config.json | Configuration | 1.5 KB | — |
| merged/generation_config.json | Configuration | 238 B | — |
| run_info.json | Configuration | 720 B | — |
| trainer/checkpoint-100/adapter_config.json | Configuration | 1.3 KB | — |
| trainer/checkpoint-100/trainer_state.json | Configuration | 96.4 KB | — |
| trainer/checkpoint-200/adapter_config.json | Configuration | 1.3 KB | — |
| trainer/checkpoint-200/trainer_state.json | Configuration | 192.3 KB | — |
| trainer/checkpoint-300/adapter_config.json | Configuration | 1.3 KB | — |
| trainer/checkpoint-300/trainer_state.json | Configuration | 288.2 KB | — |
| README.md | Documentation | 1.5 KB | — |
| lora/README.md | Documentation | 5.2 KB | — |
| trainer/README.md | Documentation | 1.9 KB | — |
| trainer/checkpoint-100/README.md | Documentation | 5.2 KB | — |
| trainer/checkpoint-200/README.md | Documentation | 5.2 KB | — |
| trainer/checkpoint-300/README.md | Documentation | 5.2 KB | — |
| merged/chat_template.jinja | Other | 2.5 KB | — |
| trainer/checkpoint-100/chat_template.jinja | Other | 2.5 KB | — |
| trainer/checkpoint-200/chat_template.jinja | Other | 2.5 KB | — |
| trainer/checkpoint-300/chat_template.jinja | Other | 2.5 KB | — |
| trainer/runs/Sep20_17-38-39_autodl-container-denn6q3zb9-b5ae9846/events.out.tfevents.1789897119.autodl-container-denn6q3zb9-b5ae9846.15373.0 | Other | 460.5 KB | 836bb9a3bbf7 |
| .gitattributes | Repository | 1.8 KB | — |
| merged/tokenizer.json | Tokenizer | 11.4 MB | 813f14d249e5 |
| merged/tokenizer_config.json | Tokenizer | 7.4 KB | — |
| trainer/checkpoint-100/tokenizer.json | Tokenizer | 11.4 MB | 813f14d249e5 |
| trainer/checkpoint-100/tokenizer_config.json | Tokenizer | 4.7 KB | — |
| trainer/checkpoint-200/tokenizer.json | Tokenizer | 11.4 MB | 813f14d249e5 |
| trainer/checkpoint-200/tokenizer_config.json | Tokenizer | 4.7 KB | — |
| trainer/checkpoint-300/tokenizer.json | Tokenizer | 11.4 MB | 813f14d249e5 |
| trainer/checkpoint-300/tokenizer_config.json | Tokenizer | 4.7 KB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 3.8 GB
Released by Rain through its official repository on Hugging Face.
Built From
- Derived from Qwen/Qwen2.5-Coder-1.5B-Instruct
- Described by arXiv:2305.03111
- Trained on (disclosed) birdsql/bird23-train-filtered
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.8 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.