Open-weight model
Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA
by Sam EnMin Huang tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA
Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA is an open-weight model from Sam EnMin Huang, released under other. Its published files total 502.1 MB.
Two 50-step LoRA r32 stages (s1, s2) on top of tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1 with a re-labelled difficulty schedule over a 64.7k-prompt pool (MATH-12k + DeepMath + DAPO-17k + DeepScaleR), CISPO, AdamW 2e-5.
Model Card
Two 50-step LoRA r32 stages (s1, s2) on top of tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1 with a re-labelled difficulty schedule over a 64.7k-prompt pool (MATH-12k + DeepMath + DAPO-17k + DeepScaleR), CISPO, AdamW 2e-5. Result: flat on MATH-500 (levels 3-5 55.0 / 55.3 vs 55.2) although the schedule reduced dead prompt groups from 47% to 6-9%; kept for the record. s2/finaladapter applies to the merge of s1, not to v1 directly. Details: https://github.com/TBOO-Y/cs2881r-sheldon-sft (results/rlvr/NOTES.md).
Excerpt from the card by Sam EnMin Huang, licensed other.
Identity and Version
- Repository
- tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA
- Publisher
- Sam EnMin Huang
- Task
- Not stated by the source
- Modality
- Other
- Library
- peft
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- dd7f90c1e9ee6b43be930432921d3b274f2b60c9
- First published
- 2026-09-27
- Last updated
- 2026-09-27
Files and Weights
20 files, 502.1 MB in total. The weights are 4 files totalling 479.1 MB in bin, safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| s1/final_adapter/adapter_model.safetensors | Weights | 239.5 MB | a2c6a6666eff |
| s1/final_adapter/training_args.bin | Weights | 7.8 KB | 3460efb2f7af |
| s2/final_adapter/adapter_model.safetensors | Weights | 239.5 MB | 021a2b429e9e |
| s2/final_adapter/training_args.bin | Weights | 7.8 KB | dd979eb3833a |
| s1/final_adapter/adapter_config.json | Configuration | 1.2 KB | — |
| s1/rlvr_args.json | Configuration | 1.2 KB | — |
| s1/trainer_state.json | Configuration | 90.0 KB | — |
| s2/final_adapter/adapter_config.json | Configuration | 1.2 KB | — |
| s2/rlvr_args.json | Configuration | 1.2 KB | — |
| s2/trainer_state.json | Configuration | 89.6 KB | — |
| README.md | Documentation | 875 B | — |
| s1/final_adapter/README.md | Documentation | 5.2 KB | — |
| s2/final_adapter/README.md | Documentation | 5.2 KB | — |
| s1/final_adapter/chat_template.jinja | Other | 2.5 KB | — |
| s2/final_adapter/chat_template.jinja | Other | 2.5 KB | — |
| .gitattributes | Repository | 1.7 KB | — |
| s1/final_adapter/tokenizer.json | Tokenizer | 11.4 MB | 3fd169731d2c |
| s1/final_adapter/tokenizer_config.json | Tokenizer | 693 B | — |
| s2/final_adapter/tokenizer.json | Tokenizer | 11.4 MB | 3fd169731d2c |
| s2/final_adapter/tokenizer_config.json | Tokenizer | 693 B | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 479.1 MB
Released by Sam EnMin Huang through its official repository on Hugging Face. Read the license.
Built From
- Adapter of tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1
- Derived from tbooy/Qwen2.5-3B-Instruct-Sheldon-RLVR-math-v1
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 479.1 MB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA
What license is Qwen2.5-3B-Instruct-Sheldon-RLVR-math-3b-LoRA released under?
other, as its publisher declares it. Read the license text before commercial use.