Downloads · 30 days
0
OPSD-Qwen3.5-9B-LoRA-strict-iter27
LSW142857/OPSD-Qwen3.5-9B-LoRA-strict-iter27
OPSD-Qwen3.5-9B-LoRA-strict-iter27 is a machine learning model from LSW142857. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for other.
Public adapter-only snapshot of the latest dated strict OPSDbaseline run (2026-08-10). The Qwen3.5-9B base model is intentionally not included.
Access
Public
Updated Aug 16, 2026
Repo size
216 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors216 MB · 100%
At a glance
- Library
- other
- Access
- Public
- Created
- Aug 16, 2026
- Updated
- Aug 16, 2026
- SHA
- bc30b464
Base models
Model card
From the Hugging Face model README
Qwen3.5-9B OPSD strict LoRA adapter — iteration 27
Public adapter-only snapshot of the latest dated strict OPSD_baseline run
(2026-08-10). The Qwen3.5-9B base model is intentionally not included.
Adapter format
This is a raw Slime/Megatron LoRA safetensors adapter, not a PEFT directory. It contains 168 BF16 LoRA tensors and must be loaded with:
- base model:
Qwen/Qwen3.5-9B - LoRA rank:
64 - LoRA alpha:
128 - dropout:
0.0 - scaling:
2.0
SHA-256:
7fec4495cc0cc4beb0583a9d6ebb39be50abc51125c63736a100eb1f1aa17e52
Download
hf download LSW142857/OPSD-Qwen3.5-9B-LoRA-strict-iter27 \
adapter/checkpoint_iter_0000027.safetensors \
--repo-type model \
--local-dir .
hf download LSW142857/OPSD-Qwen3.5-9B-LoRA-strict-iter27 \
adapter/checkpoint_iter_0000027.manifest.json \
--repo-type model \
--local-dir .
Evaluation
Use the adapter with the same OPSD_baseline checkout and a local
Qwen3.5-9B base model:
QWEN_LORA_RANK=64 QWEN_LORA_ALPHA=128 \
bash examples/coding_agent_rl/evaluate_qwen35_9b_opsd_lora.sh \
adapter/checkpoint_iter_0000027.safetensors \
/path/to/Qwen3.5-9B \
/path/to/eval.jsonl
The adapter does not contain optimizer state, the base model, SWE-Gym images, or training data.
- Library
- other
- Created
- Aug 16, 2026
- Updated
- Aug 16, 2026