From: Aman Gupta <redacted>
Date: Mon, 16 Mar 2026 14:07:13 +0000 (+0800)
Subject: llama-graph: replace cont with reshape for alpha in qwen35 (#20640)
X-Git-Tag: upstream/0.0.8611~234
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=3c8521c4f59a07302076dc5d41424bb3a25777d8;p=pkg%2Fggml%2Fsources%2Fllama.cpp

llama-graph: replace cont with reshape for alpha in qwen35 (#20640)
---

diff --git a/src/models/qwen35.cpp b/src/models/qwen35.cpp
index 3108bf331..d07579ee8 100644
--- a/src/models/qwen35.cpp
+++ b/src/models/qwen35.cpp
@@ -224,7 +224,7 @@ ggml_tensor * llm_build_qwen35::build_layer_attn_linear(
     beta = ggml_sigmoid(ctx0, beta);
 
     ggml_tensor * alpha = build_lora_mm(model.layers[il].ssm_alpha, cur, model.layers[il].ssm_alpha_s);
-    alpha = ggml_cont_3d(ctx0, alpha, num_v_heads, n_seq_tokens, n_seqs);
+    alpha = ggml_reshape_3d(ctx0, alpha, num_v_heads, n_seq_tokens, n_seqs);
     cb(alpha, "alpha", il);
 
     ggml_tensor * alpha_biased   = ggml_add(ctx0, alpha, model.layers[il].ssm_dt);
diff --git a/src/models/qwen35moe.cpp b/src/models/qwen35moe.cpp
index 165e2412e..b38660c0b 100644
--- a/src/models/qwen35moe.cpp
+++ b/src/models/qwen35moe.cpp
@@ -224,7 +224,7 @@ ggml_tensor * llm_build_qwen35moe ::build_layer_attn_linear(
     beta = ggml_sigmoid(ctx0, beta);
 
     ggml_tensor * alpha = build_lora_mm(model.layers[il].ssm_alpha, cur, model.layers[il].ssm_alpha_s);
-    alpha = ggml_cont_3d(ctx0, alpha, num_v_heads, n_seq_tokens, n_seqs);
+    alpha = ggml_reshape_3d(ctx0, alpha, num_v_heads, n_seq_tokens, n_seqs);
     cb(alpha, "alpha", il);
 
     ggml_tensor * alpha_biased   = ggml_add(ctx0, alpha, model.layers[il].ssm_dt);