shawn zhang

shawnzzzzz's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

updated a model 27 days ago

shawnzzzzz/Qwen3-30B-A3B-GRPO-W4A4-NoOverlong-Step800

Reinforcement Learning • 31B • Updated 27 days ago • 87

published a model 27 days ago

shawnzzzzz/Qwen3-30B-A3B-GRPO-W4A4-NoOverlong-Step800

Reinforcement Learning • 31B • Updated 27 days ago • 87

updated 3 models about 1 month ago

shawnzzzzz/Qwen3-30B-A3B-GRPO-3Loss-Step800

Text Generation • 31B • Updated Jul 27 • 46

shawnzzzzz/Qwen3-30B-A3B-GRPO-BF16-Control-Step800

Text Generation • 31B • Updated Jul 27 • 36

shawnzzzzz/Qwen3-30B-A3B-GRPO-LowPrecision-Control-Step410

Text Generation • 31B • Updated Jul 27 • 34

published 3 models about 1 month ago

shawnzzzzz/Qwen3-30B-A3B-GRPO-LowPrecision-Control-Step410

Text Generation • 31B • Updated Jul 27 • 34

shawnzzzzz/Qwen3-30B-A3B-GRPO-3Loss-Step800

Text Generation • 31B • Updated Jul 27 • 46

shawnzzzzz/Qwen3-30B-A3B-GRPO-BF16-Control-Step800

Text Generation • 31B • Updated Jul 27 • 36

updated a collection about 2 months ago

Qwen3-30B-A3B W4A4-QAT vs BF16 Checkpoints

Collection

Seven step-matched Qwen3-30B-A3B DAPO pairs: FFN W4A4-QAT BF16 master weights (not packed W4A4) and a BF16 baseline. • 14 items • Updated Jul 12