Sub-3-bit Qwen3 recovered with quantization-aware distillation then on-policy distillation. -QAD entries are training starting points.
Chen Yuan
MingZwhy
AI & ML interests
None yet
Recent Activity
updated a collection 3 days ago
QAOPD updated a collection 3 days ago
QAOPD updated a collection 3 days ago
QAOPDOrganizations
models 13
MingZwhy/Qwen3-4B-W1.88-QAD
Text Generation • 196k • Updated • 133
MingZwhy/Qwen3-4B-W2.79-QAD
Text Generation • 196k • Updated • 135
MingZwhy/Qwen3-1.7B-W1.88-QAD
Text Generation • 2B • Updated • 134
MingZwhy/Qwen3-1.7B-W2.79-QAD
Text Generation • 2B • Updated • 136
MingZwhy/Qwen3-0.6B-W1.88-QAD
Text Generation • 0.6B • Updated • 264
MingZwhy/Qwen3-0.6B-W2.79-QAD
Text Generation • 0.6B • Updated • 263
MingZwhy/Qwen3-4B-W1.88-QAOPD
Text Generation • 4B • Updated • 244
MingZwhy/Qwen3-4B-W2.79-QAOPD
Text Generation • 4B • Updated • 225
MingZwhy/Qwen3-1.7B-W1.88-QAOPD
Text Generation • 2B • Updated • 204
MingZwhy/Qwen3-1.7B-W2.79-QAOPD
Text Generation • 2B • Updated • 191
datasets 0
None public yet