← Back to Quantization

int6 per-row with GPTQ Hessian-aware quantization

Quantization
Used in
1 PRs
Best BPB
1.1355
Avg BPB
1.1355

Submissions

PR #579by newjordan
1.1355

Hyperparameters Across PRs

pr_numberbitsscope
5796MLP and attention weights