llama : fix llama3.1 rope_freqs not respecting custom head_dim (#9141)

* fix: llama3.1 rope_freqs not respecting custom head_dim * fix: use potential head_dim for Exaone
ggerganov · Aug 27, 2024 · 75e1dbb · 75e1dbb
1 parent ad76569
commit 75e1dbb
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 4 deletions.
diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
@@ -1572,7 +1572,7 @@ def prepare_tensors(self):
         if rope_scaling := self.find_hparam(["rope_scaling"], optional=True):
             if rope_scaling.get("rope_type", '').lower() == "llama3":
                 base = self.hparams.get("rope_theta", 10000.0)
-                dim = self.hparams["hidden_size"] // self.hparams["num_attention_heads"]
+                dim = self.hparams.get("head_dim", self.hparams["hidden_size"] // self.hparams["num_attention_heads"])
                 freqs = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
 
                 factor = rope_scaling.get("factor", 8.0)
@@ -3820,7 +3820,7 @@ def prepare_tensors(self):
         if rope_scaling := self.find_hparam(["rope_scaling"], optional=True):
             if rope_scaling.get("rope_type", '').lower() == "llama3":
                 base = self.hparams.get("rope_theta", 10000.0)
-                dim = self.hparams["hidden_size"] // self.hparams["num_attention_heads"]
+                dim = self.hparams.get("head_dim", self.hparams["hidden_size"] // self.hparams["num_attention_heads"])
                 freqs = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
 
                 factor = rope_scaling.get("factor", 8.0)

diff --git a/src/llama.cpp b/src/llama.cpp
@@ -6605,6 +6605,7 @@ static bool llm_load_tensors(
         const int64_t n_embd_gqa    = n_embd_v_gqa;
         const int64_t n_vocab       = hparams.n_vocab;
         const int64_t n_vocab_type  = hparams.n_vocab_type;
+        const int64_t n_rot         = hparams.n_rot;
         const int64_t n_expert      = hparams.n_expert;
         const int64_t n_expert_used = hparams.n_expert_used;
         const int64_t n_ctx_train   = hparams.n_ctx_train;
@@ -6662,7 +6663,7 @@ static bool llm_load_tensors(
 
                         layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
 
-                        layer.rope_freqs = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ROPE_FREQS, "weight"), {n_embd/n_head/2}, llama_model_loader::TENSOR_NOT_REQUIRED | (i != 0 ? llama_model_loader::TENSOR_DUPLICATED : 0));
+                        layer.rope_freqs = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ROPE_FREQS, "weight"), {n_rot/2}, llama_model_loader::TENSOR_NOT_REQUIRED | (i != 0 ? llama_model_loader::TENSOR_DUPLICATED : 0));
 
                         if (n_expert == 0) {
                             layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff});
@@ -8193,7 +8194,7 @@ static bool llm_load_tensors(
                         layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd});
 
                         layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
-                        layer.rope_freqs = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ROPE_FREQS, "weight"), {n_embd/n_head/2}, llama_model_loader::TENSOR_NOT_REQUIRED | (i != 0 ? llama_model_loader::TENSOR_DUPLICATED : 0));
+                        layer.rope_freqs = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ROPE_FREQS, "weight"), {n_rot/2}, llama_model_loader::TENSOR_NOT_REQUIRED | (i != 0 ? llama_model_loader::TENSOR_DUPLICATED : 0));
                         layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd,   n_ff});
                         layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  n_ff, n_embd});
                         layer.ffn_up   = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff});