Skip to content

Commit

Permalink
Enable noise tokens for Qwen2-VL quantizatino
Browse files Browse the repository at this point in the history
  • Loading branch information
turboderp committed Dec 7, 2024
1 parent c55656c commit ba9774f
Showing 1 changed file with 2 additions and 0 deletions.
2 changes: 2 additions & 0 deletions exllamav2/architecture.py
Original file line number Diff line number Diff line change
Expand Up @@ -402,6 +402,8 @@ class Params:
self.mmp.mlp_bias = True
self.mmp.norm = "layernorm"

self.standard_calib_noise = (5, 30)

# Gemma

if arch_string == "GemmaForCausalLM":
Expand Down

0 comments on commit ba9774f

Please sign in to comment.