HyperFlow H3 Flash: low-rank down factors quantized to INT8 ConvRot (smaller files, faster steps); NOTICE updated

HyperFlow-H3-Flash-NOTICE

@@ -25,4 +25,5 @@ MiniMax-H3-Ref2VA-HyperFlow-Flash-14B_int8_convrot.safetensors are fl2va_int8_co
 https://huggingface.co/VideoRebirthAI/hyperflow-h3-flash (revision 9e9ed14b8be4b751d7517b8ca58dd199d509720c), each merged into
 one file. Modifications: tensor keys renamed to WanGP's MiniMax H3 names; the two row halves of each feed-forward input
 projection swapped (gate | value order); the INT8 weights and their scales stored as WanGP INT8 ConvRot tensors; the RoPE
-frequency buffer added. Tensor values are unchanged. The Agreement mentioned above is included here as HyperFlow-H3-Flash-LICENSE.
+frequency buffer added; the low-rank down factors, stored in BF16 upstream, quantized to INT8 ConvRot (per-row
+scales, Hadamard groups of 256). Other tensor values are unchanged. The Agreement mentioned above is included here as HyperFlow-H3-Flash-LICENSE.

MiniMax-H3-FL2VA-HyperFlow-Flash-14B_int8_convrot.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b3f9d8f5afac040db1bea831480c4e5db3fd5085ce24c6af271bc6952b809e61
-size 15880662768
+oid sha256:736c943254732bcd621b43c91181184ea2538f9033288543fbf9cfefe85ab8cb
+size 15272470528

MiniMax-H3-Ref2VA-HyperFlow-Flash-14B_int8_convrot.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:362a043209b312d22198cd72d1091d9529d43c75df5ef9134c986e44d0394282
-size 16311238704
+oid sha256:d1c046a3b51b6e0413ad836d33afe2fd363378103d391e51928a5bbd424d71a4
+size 15598167356