Downloads · 30 days
171
21% of all-time downloads
qwen3coder-calendar-if-v49-lr2-step3
penfever/qwen3coder-calendar-if-v49-lr2-step3
qwen3coder-calendar-if-v49-lr2-step3 is a text generation model from penfever. Use it when you need the model to write or continue text. It is set up for transformers.
Qwen3-Coder-30B-A3B-Instruct after asynchronous RLOO training on the TaskTrove v4.9 calendar instruction-following source. This checkpoint scored 0.296875 mean reward (38/128 passes) on the fixed holdout. The holdout…
All-time downloads
799
Public
Parameters
30.5B
61.1 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors61.1 GB · 100%
At a glance
- Task
- Text Generation
- Library
- transformers
- Model type
- qwen3_moe
- Access
- Public
- Created
- Sep 7, 2026
- Updated
- Sep 7, 2026
- SHA
- bda2eadd
Try a prompt
Base models
Model card
From the Hugging Face model README
Qwen3-Coder calendar instruction-following RL — step 3
Qwen3-Coder-30B-A3B-Instruct after asynchronous RLOO training on the TaskTrove v4.9 calendar instruction-following source. This checkpoint scored 0.296875 mean reward (38/128 passes) on the fixed holdout. The holdout was later found to overlap the training source, so this score is a selection statistic, not an uncontaminated estimate of generalization.
Native checkpoint: q3c-rl-calendar-if-v49-nodapo-lr2-r1/global_step_3.
Training Traces
The complete experiment artifacts, configs, metrics, trace samples, tracker, and contamination analysis are in qwen3coder-iris-rl-data-sweep-artifacts.
- Task
- Text Generation
- Library
- transformers
- Type
- qwen3_moe
- Created
- Sep 7, 2026
- Updated
- Sep 7, 2026