gguf : get rid of n_mult, read n_ff from file

gguf-llama.cpp

@@ -177,15 +177,12 @@ struct llama_hparams {
     uint32_t n_vocab   = 32000;
     uint32_t n_ctx     = 512;   // this is provided as user input?
     uint32_t n_embd    = 4096;
-    uint32_t n_mult    = 256;
     uint32_t n_head    = 32;
     uint32_t n_head_kv = 32;
     uint32_t n_layer   = 32;
     uint32_t n_rot     = 64;
+    uint32_t n_ff       = 11008;
 
-    // LLaMAv2
-    // TODO: load from model data hparams
-    float f_ffn_mult = 1.0f;
     float f_rms_norm_eps = LLAMA_DEFAULT_RMS_EPS;
 
     float rope_freq_base  = 10000.0f;
@@ -467,7 +464,7 @@ struct llama_load_tensors_map {
 };
 
 enum gguf_file_version {
-    GGUF_FILE_VERSION_V1
+    GGUF_FILE_VERSION_V1 = 1,
     
 };
 
@@ -490,6 +487,7 @@ struct ggml_context * ctx_data = NULL;
     };
 
     gguf_ctx = gguf_init_from_file(fname, params);
+    file_version = (enum gguf_file_version) gguf_get_version(gguf_ctx);
 
     read_hparams();
     read_vocab();
@@ -505,6 +503,15 @@ struct ggml_context * ctx_data = NULL;
         return gguf_get_val_u32(gguf_ctx, i);
     }
 
+    float read_f32(const char * key) {
+        int i = gguf_find_key(gguf_ctx, key);
+        if (i == -1) {
+            throw std::runtime_error(format("cannot find param with key %s\n", key));
+        }
+
+        return gguf_get_val_f32(gguf_ctx, i);
+    }
+
     int read_n_vocab() {
         int i = gguf_find_key(gguf_ctx, "tokenizer.ggml.tokens");
     if (i == -1) {
@@ -514,18 +521,6 @@ struct ggml_context * ctx_data = NULL;
     return gguf_get_arr_n(gguf_ctx, i);
     }
 
-    int find_n_mult(const int n_ff, const int n_embd) {
-        int n_mults[3] = {8192, 1, -1};
-        for (int i = 0; i < 3; ++i) {
-            int calc_ff = (((8 * n_embd) / 3 + n_mults[i] - 1) / n_mults[i]) * n_mults[i];
-            if (calc_ff == n_ff) {
-                return n_mults[i];
-            }
-        }
-
-        throw std::runtime_error(format("failed to find n_mult for n_ff = %d and n_embd = %d\n", n_ff, n_embd));
-    }
-
     void read_hparams() {
 
         // TODO make keysconstants in header
@@ -533,13 +528,11 @@ struct ggml_context * ctx_data = NULL;
         hparams.n_vocab = read_n_vocab();
         hparams.n_ctx   = read_u32("llama.context_length");
         hparams.n_embd  = read_u32("llama.embedding_length");
-        uint32_t n_ff    = read_u32("llama.feed_forward_length");
+        hparams.n_ff    = read_u32("llama.feed_forward_length");
-        GGML_UNUSED(n_ff);
-        //hparams.n_mult  = find_n_mult(n_ff, hparams.n_embd);
         hparams.n_head  = read_u32("llama.attention.head_count");
         hparams.n_layer = read_u32("llama.layer_count");
-        hparams.n_rot   = hparams.n_embd / hparams.n_head;
+        hparams.n_rot   = read_u32("llama.rope.dimension_count");
-        //hparams.ftype   = (enum llama_ftype) file.read_u32();
+        hparams.f_rms_norm_eps = read_f32("llama.attention.layer_norm_rms_epsilon");
         
         // LLaMAv2
         // hparams.n_head_kv = read_u32("llama.attention.head_count_kv");
@@ -1125,6 +1118,7 @@ static void llama_model_load_internal(
         bool vocab_only,
         llama_progress_callback progress_callback,
         void * progress_callback_user_data) {
+            GGML_UNUSED(rms_norm_eps); // TODO: update function signature to remove this
 
     model.t_start_us = ggml_time_us();
 
@@ -1137,9 +1131,6 @@ static void llama_model_load_internal(
 
     auto & hparams = model.hparams;
 
-    // TODO: read from file
-    hparams.f_rms_norm_eps = rms_norm_eps;
-
     {
         switch (hparams.n_layer) {
             case 26: model.type = e_model::MODEL_3B; break;
@@ -1162,25 +1153,19 @@ static void llama_model_load_internal(
         if (model.type == e_model::MODEL_65B && n_gqa == 8) {
             fprintf(stderr, "%s: warning: assuming 70B model based on GQA == %d\n", __func__, n_gqa);
             model.type = e_model::MODEL_70B;
-            hparams.f_ffn_mult = 1.3f; // from the params.json of the 70B model
             }
 
         hparams.rope_freq_base  = rope_freq_base;
         hparams.rope_freq_scale = rope_freq_scale;
     }
 
-    // ref: https://github.com/facebookresearch/llama/blob/6c7fe276574e78057f917549435a2554000a876d/llama/model.py#L194-L199
+    const uint32_t n_ff = hparams.n_ff;
-    const uint32_t n_ff_raw  = 2*(4*hparams.n_embd)/3;
-    const uint32_t n_ff_mult = hparams.f_ffn_mult*n_ff_raw;
-    const uint32_t n_ff      = ((n_ff_mult + hparams.n_mult - 1)/hparams.n_mult)*hparams.n_mult;
-    //const uint32_t n_ff = 28672;
     
     {
         fprintf(stderr, "%s: format     = %s\n",   __func__, gguf_file_version_name(file_version));
         fprintf(stderr, "%s: n_vocab    = %u\n",   __func__, hparams.n_vocab);
         fprintf(stderr, "%s: n_ctx      = %u\n",   __func__, hparams.n_ctx);
         fprintf(stderr, "%s: n_embd     = %u\n",   __func__, hparams.n_embd);
-        fprintf(stderr, "%s: n_mult     = %u\n",   __func__, hparams.n_mult);
         fprintf(stderr, "%s: n_head     = %u\n",   __func__, hparams.n_head);
         fprintf(stderr, "%s: n_head_kv  = %u\n",   __func__, hparams.n_head_kv);
         fprintf(stderr, "%s: n_layer    = %u\n",   __func__, hparams.n_layer);