llama : fix n_rot default (ggerganov#8348)

ggml-ci
Nexesenex · Jul 11, 2024 · 1bf7194 · 1bf7194
1 parent 6030c8b
commit 1bf7194
Showing 1 changed file with 15 additions and 14 deletions.
diff --git a/llama.cpp b/llama.cpp
@@ -4745,16 +4745,6 @@ static void llm_load_hparams(
 
     // non-transformer models do not have attention heads
     if (hparams.n_head() > 0) {
-        // sanity check for n_rot (optional)
-        hparams.n_rot = hparams.n_embd / hparams.n_head();
-
-        ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT, hparams.n_rot, false);
-
-        if (model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON) {
-            if (hparams.n_rot != hparams.n_embd / hparams.n_head()) {
-                throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd / hparams.n_head()));
-            }
-        }
         // gpt-neox n_rot = rotary_pct * (n_embd / n_head)
         // gpt-j n_rot = rotary_dim
 
@@ -4763,6 +4753,17 @@ static void llm_load_hparams(
 
         hparams.n_embd_head_v = hparams.n_embd / hparams.n_head();
         ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH, hparams.n_embd_head_v, false);
+
+        // sanity check for n_rot (optional)
+        hparams.n_rot = hparams.n_embd_head_k;
+
+        ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT, hparams.n_rot, false);
+
+        if (model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON) {
+            if (hparams.n_rot != hparams.n_embd_head_k) {
+                throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd_head_k));
+            }
+        }
     } else {
         hparams.n_rot = 0;
         hparams.n_embd_head_k = 0;
@@ -11650,7 +11651,7 @@ struct llm_build_context {
 
                 Qcur = ggml_rope_ext(
                         ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head,    n_tokens), inp_pos, nullptr,
-                        n_embd_head_k, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
                         ext_factor, attn_factor, beta_fast, beta_slow);
                 cb(Qcur, "Qcur", il);
 
@@ -11659,7 +11660,7 @@ struct llm_build_context {
 
                 Kcur = ggml_rope_ext(
                         ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv, n_tokens), inp_pos, nullptr,
-                        n_embd_head_k, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
                         ext_factor, attn_factor, beta_fast, beta_slow);
                 cb(Kcur, "Kcur", il);
 
@@ -11763,7 +11764,7 @@ struct llm_build_context {
 
                 Qcur = ggml_rope_ext(
                         ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head,    n_tokens), inp_pos, nullptr,
-                        n_embd_head_k, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
                         ext_factor, attn_factor, beta_fast, beta_slow);
                 cb(Qcur, "Qcur", il);
 
@@ -11772,7 +11773,7 @@ struct llm_build_context {
 
                 Kcur = ggml_rope_ext(
                         ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv, n_tokens), inp_pos, nullptr,
-                        n_embd_head_k, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
                         ext_factor, attn_factor, beta_fast, beta_slow);
                 cb(Kcur, "Kcur", il);