model: apply .scale tensors automatically + fix correctness (#32669)

* model: apply .scale tensors automatically

* rm build_moe_ffn "_s" overload

* use llama_prec_policy

* fix ci

* build_lora_mm_id: rm w_s
This commit is contained in:
Xuan-Son Nguyen authored and GitHub committed 2026-11-21 16:30:01 +01:00
1 parent e50c254a22
commit 19bc41dbb4
39 files changed
+164 -290

No files matched your search

+5
View File
@@ -113,6 +113,11 @@ void llama_model_glm4_moe::load_arch_tensors(llama_model_loader & ml) {
layer.nextn.shared_head_head = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD, "weight", i), { n_embd, n_vocab }, TENSOR_NOT_REQUIRED | flags);
layer.nextn.shared_head_norm = create_tensor(tn(LLM_TENSOR_NEXTN_SHARED_HEAD_NORM, "weight", i), { n_embd }, TENSOR_NOT_REQUIRED | flags);
}
// numerical issues with F16 accumulators
prec_policy.set_acc(layer.wo, GGML_PREC_BF16);
prec_policy.set_acc(layer.ffn_down, GGML_PREC_BF16);
prec_policy.set_acc(layer.ffn_down_shexp, GGML_PREC_BF16);
}
}