vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-06 12:40:02 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

fix AMX

Browse source
This commit is contained in:
Atream 2025-05-07 19:12:19 +08:00 • committed by GitHub
parent 3f14e311cb
commit b22cded890
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
csrc/ktransformers_ext/operators/amx/moe.hpp
View file

@ -272,8 +272,8 @@ public:
void forward(int qlen, int k, const uint64_t *expert_ids, const float *weights, const void *input, void *output, void forward(int qlen, int k, const uint64_t *expert_ids, const float *weights, const void *input, void *output,
int *batch_size_tensor, Backend *backend) { int *batch_size_tensor, Backend *backend) {
bool use_amx = (qlen > 4 * config_.expert_num / config_.routed_expert_num);
qlen = batch_size_tensor[0]; qlen = batch_size_tensor[0];
bool use_amx = (qlen > 4 * config_.expert_num / config_.routed_expert_num);
int activated_expert = 0; int activated_expert = 0;
for (int i = 0; i < config_.expert_num; i++) { for (int i = 0; i < config_.expert_num; i++) {
m_local_num_[i] = 0; m_local_num_[i] = 0;