imdatta0
qwen3-4b-swegym-moto-kl02-sft20k-mixhardx4-adapter
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Get help setting up a custom Dedicated Endpoints.
Talk with our engineer to get a quote for reserved GPU instances with discounts.
README
Qwen3-4B SWE-Gym moto KL02 mixed refresh plus hard-multi SFT adapter
This repository contains a PEFT LoRA adapter checkpoint from the Qwen3 SWE-Gym moto investigation.
Save-only SFT on top of the KL02 adapter using mixed visible16k multihint-refresh teacher data plus 20k hard-multi teacher data repeated four times. Built-in post-SFT greedy eval: 10/35, mean reward 0.4734, patch applied rate 0.8. Full pass@8 was not run before upload.
Local source checkpoint before upload:
/mnt/disks/unslothai/datta0/cache/qwen3-grpo-patch/20260605_053038_swegym_q4b-kl02-sft20k-mixhardx4_10e3a3b/checkpoints/best_holdout
Use with the base model unsloth/Qwen3-4B-Instruct-2507.
Model provider
imdatta0
Model tree
Base
unsloth/Qwen3-4B-Instruct-2507
Adapter
this model
Modalities
Input
Text
Output
Text
Pricing
Dedicated Endpoints
View detailsSupported Functionality
Model APIs
Dedicated Endpoints
Container
More information