Zhongzhu commited on
Commit
3fe792a
·
verified ·
1 Parent(s): 6cadfbe

Add MiniMax-M2.7 OSCAR KV rotations

Browse files
MiniMax-M2.7/README.md ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ # MiniMax-M2.7 OSCAR KV rotations
2
+
3
+ Shared rotation, 62 layers, `[128, 128]`, uniform quantizer with Lloyd-Max off.
4
+ This is the `seq10000_prompt37_group128` calibration, the one every published
5
+ M2.7 eval config references.
MiniMax-M2.7/k_rotation_qqt_r_h_pbr.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a753379adf04ed17f364f378967bff0a568670cbc352fd6fa4618cbd57afae3e
3
+ size 4131493
MiniMax-M2.7/v_rotation_sst_r_h_pbr.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4ef8e204a7253aa5fc845513bc3bdcbf82a13da61dc1972cf720f562ef82c9b
3
+ size 4131493