model : fix wavtokenizer embedding notions (#19479)

diff --git a/src/llama-hparams.h b/src/llama-hparams.h
index 6c695bdbf6625d3db6076995a790ec845406baa7..706eda8441bc74b54ef78eb01f88da53a21d83c3 100644 (file)
--- a/src/llama-hparams.h
+++ b/src/llama-hparams.h
@@ -42,7 +42,6 @@ struct llama_hparams {
 
     uint32_t n_ctx_train; // context size the model was trained on
     uint32_t n_embd;
-    uint32_t n_embd_features = 0;
     uint32_t n_layer;
     int32_t n_layer_kv_from_start = -1; // if non-negative, the first n_layer_kv_from_start layers have KV cache
     uint32_t n_rot;

diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 7a06e96c8796b379425a7f4329e7dab2c5ec09c1..5816e9a954cc679dffa3af2f2ec65c63811220be 100644 (file)
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -523,7 +523,8 @@ void llama_model::load_hparams(llama_model_loader & ml) {
     ml.get_key(LLM_KV_EXPERT_GROUP_USED_COUNT, hparams.n_group_used,    false);
 
     if (arch == LLM_ARCH_WAVTOKENIZER_DEC) {
-        ml.get_key(LLM_KV_FEATURES_LENGTH, hparams.n_embd_features);
+        ml.get_key(LLM_KV_FEATURES_LENGTH,  hparams.n_embd);
+        ml.get_key(LLM_KV_EMBEDDING_LENGTH, hparams.n_embd_out_impl);
 
         ml.get_key(LLM_KV_POSNET_EMBEDDING_LENGTH, hparams.posnet.n_embd);
         ml.get_key(LLM_KV_POSNET_BLOCK_COUNT,      hparams.posnet.n_layer);
@@ -6046,9 +6047,9 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                 } break;
             case LLM_ARCH_WAVTOKENIZER_DEC:
                 {
-                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {hparams.n_embd_features, n_vocab}, 0);
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {hparams.n_embd, n_vocab}, 0);
 
-                    conv1d   = create_tensor(tn(LLM_TENSOR_CONV1D, "weight"), {7, hparams.n_embd_features, hparams.posnet.n_embd}, 0);
+                    conv1d   = create_tensor(tn(LLM_TENSOR_CONV1D, "weight"), {7, hparams.n_embd, hparams.posnet.n_embd}, 0);
                     conv1d_b = create_tensor(tn(LLM_TENSOR_CONV1D, "bias"),   {1, hparams.posnet.n_embd}, 0);
 
                     // posnet
@@ -6144,8 +6145,8 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         output_norm_b = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd}, 0);
                     }
 
-                    output   = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {hparams.convnext.n_embd, n_embd}, 0);
-                    output_b = create_tensor(tn(LLM_TENSOR_OUTPUT, "bias"),   {n_embd}, 0);
+                    output   = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {hparams.convnext.n_embd, hparams.n_embd_out()}, 0);
+                    output_b = create_tensor(tn(LLM_TENSOR_OUTPUT, "bias"),   {hparams.n_embd_out()}, 0);
                 } break;
             case LLM_ARCH_BAILINGMOE:
                 {

diff --git a/tools/tts/tts.cpp b/tools/tts/tts.cpp
index 8c39fce8baad76af7889e8b051631f8d0c1cea78..ac55a8b1ca41ad766d6f619371761e9891a0a26a 100644 (file)
--- a/tools/tts/tts.cpp
+++ b/tools/tts/tts.cpp
@@ -1036,7 +1036,7 @@ lovely<|t_0.56|><|code_start|><|634|><|596|><|1766|><|1556|><|1306|><|1285|><|14
 
 #if 1
     // spectral operations
-    const int n_embd = llama_model_n_embd(model_cts);
+    const int n_embd = llama_model_n_embd_out(model_cts);
     const float * embd = llama_get_embeddings(ctx_cts);
 
     auto audio = embd_to_audio(embd, n_codes, n_embd, params.cpuparams.n_threads);