qwen4exp : fix graph reallocation from the n-gram table scale (#32483)

* qwen4exp : apply the n-gram table scale with ggml_scale

* read the scale once after load

* remove comment
This commit is contained in:
Aman Gupta authored and GitHub committed 2026-11-18 00:35:54 +08:00
1 parent 955c4e8ee5
commit ad9d9e8f50
3 files changed
+9 -1

No files matched your search

+4
View File
@@ -2185,6 +2185,10 @@ bool llama_model_base::load_tensors(llama_model_loader & ml) {
}
}
if (per_layer_tok_embd_s) {
ggml_backend_tensor_get(per_layer_tok_embd_s, &per_layer_tok_embd_scale, 0, sizeof(per_layer_tok_embd_scale));
}
load_tensors_hadamard();
return true;