diff --git a/convert.py b/convert.py
index 8f4f0399e..ece5a0266 100644
--- a/convert.py
+++ b/convert.py
@@ -121,7 +121,6 @@ def make_tensors_list() -> List[str]:
             f'layers.{i}.feed_forward.w1.weight',
             f'layers.{i}.feed_forward.w2.weight',
             f'layers.{i}.feed_forward.w3.weight',
-            f'layers.{i}.atttention_norm.weight',
             f'layers.{i}.ffn_norm.weight',
         ]
     return ret
@@ -1055,7 +1054,7 @@ def load_some_model(path: Path) -> ModelPlus:
         files = list(path.glob("model-00001-of-*.safetensors"))
         if not files:
             # Try the PyTorch patterns too, with lower priority
-            globs = ["consolidated.00.pth", "pytorch_model-00001-of-*.bin", "*.pt"]
+            globs = ["consolidated.00.pth", "pytorch_model-00001-of-*.bin", "*.pt", "pytorch_model.bin" ]
             files = [file for glob in globs for file in path.glob(glob)]
         if not files:
             # Try GGML too, but with lower priority, since if both a non-GGML
diff --git a/examples/common.h b/examples/common.h
index 717838f06..f4e07a252 100644
--- a/examples/common.h
+++ b/examples/common.h
@@ -45,7 +45,7 @@ struct gpt_params {
     float   mirostat_tau      = 5.00f; // target entropy
     float   mirostat_eta      = 0.10f; // learning rate
 
-    std::string model  = "models/lamma-7B/ggml-model.bin"; // model path
+    std::string model  = "models/7B/ggml-model.bin"; // model path
     std::string prompt = "";
     std::string path_prompt_cache = "";  // path to file for saving/loading prompt eval state
     std::string input_prefix      = "";  // string to prefix user inputs with
diff --git a/examples/embedding/embedding.cpp b/examples/embedding/embedding.cpp
index bb3fd50a9..c24f7f820 100644
--- a/examples/embedding/embedding.cpp
+++ b/examples/embedding/embedding.cpp
@@ -6,7 +6,6 @@
 
 int main(int argc, char ** argv) {
     gpt_params params;
-    params.model = "models/llama-7B/ggml-model.bin";
 
     if (gpt_params_parse(argc, argv, params) == false) {
         return 1;
diff --git a/examples/main/main.cpp b/examples/main/main.cpp
index 8543414dd..fe1c847a7 100644
--- a/examples/main/main.cpp
+++ b/examples/main/main.cpp
@@ -50,7 +50,6 @@ void sigint_handler(int signo) {
 
 int main(int argc, char ** argv) {
     gpt_params params;
-    params.model = "models/llama-7B/ggml-model.bin";
 
     if (gpt_params_parse(argc, argv, params) == false) {
         return 1;
diff --git a/examples/perplexity/perplexity.cpp b/examples/perplexity/perplexity.cpp
index 9212dee5c..9d38626cb 100644
--- a/examples/perplexity/perplexity.cpp
+++ b/examples/perplexity/perplexity.cpp
@@ -116,7 +116,6 @@ void perplexity(llama_context * ctx, const gpt_params & params) {
 
 int main(int argc, char ** argv) {
     gpt_params params;
-    params.model = "models/llama-7B/ggml-model.bin";
 
     params.n_batch = 512;
     if (gpt_params_parse(argc, argv, params) == false) {
diff --git a/examples/save-load-state/save-load-state.cpp b/examples/save-load-state/save-load-state.cpp
index ea0a984d9..355969579 100644
--- a/examples/save-load-state/save-load-state.cpp
+++ b/examples/save-load-state/save-load-state.cpp
@@ -8,7 +8,6 @@
 
 int main(int argc, char ** argv) {
     gpt_params params;
-    params.model = "models/llama-7B/ggml-model.bin";
     params.seed = 42;
     params.n_threads = 4;
     params.repeat_last_n = 64;
diff --git a/ggml.c b/ggml.c
index d4235545b..9434c8671 100644
--- a/ggml.c
+++ b/ggml.c
@@ -541,12 +541,7 @@ static inline __m256 sum_i16_pairs_float(const __m256i x) {
     return _mm256_cvtepi32_ps(summed_pairs);
 }
 
-// multiply int8_t, add results pairwise twice and return as float vector
-static inline __m256 mul_sum_i8_pairs_float(const __m256i x, const __m256i y) {
-    // Get absolute values of x vectors
-    const __m256i ax = _mm256_sign_epi8(x, x);
-    // Sign the values of the y vectors
-    const __m256i sy = _mm256_sign_epi8(y, x);
+static inline __m256 mul_sum_us8_pairs_float(const __m256i ax, const __m256i sy) {
 #if __AVXVNNI__
     const __m256i zero = _mm256_setzero_si256();
     const __m256i summed_pairs = _mm256_dpbusd_epi32(zero, ax, sy);
@@ -558,6 +553,21 @@ static inline __m256 mul_sum_i8_pairs_float(const __m256i x, const __m256i y) {
 #endif
 }
 
+// multiply int8_t, add results pairwise twice and return as float vector
+static inline __m256 mul_sum_i8_pairs_float(const __m256i x, const __m256i y) {
+#if __AVXVNNIINT8__
+    const __m256i zero = _mm256_setzero_si256();
+    const __m256i summed_pairs = _mm256_dpbssd_epi32(zero, x, y);
+    return _mm256_cvtepi32_ps(summed_pairs);
+#else
+    // Get absolute values of x vectors
+    const __m256i ax = _mm256_sign_epi8(x, x);
+    // Sign the values of the y vectors
+    const __m256i sy = _mm256_sign_epi8(y, x);
+    return mul_sum_us8_pairs_float(ax, sy);
+#endif
+}
+
 static inline __m128i packNibbles( __m256i bytes )
 {
     // Move bits within 16-bit lanes from 0000_abcd_0000_efgh into 0000_0000_abcd_efgh
@@ -617,6 +627,17 @@ static inline __m256 sum_i16_pairs_float(const __m128i xh, const __m128i xl) {
     return _mm256_cvtepi32_ps(summed_pairs);
 }
 
+static inline __m256 mul_sum_us8_pairs_float(const __m256i ax, const __m256i sy) {
+    const __m128i axl = _mm256_castsi256_si128(ax);
+    const __m128i axh = _mm256_extractf128_si256(ax, 1);
+    const __m128i syl = _mm256_castsi256_si128(sy);
+    const __m128i syh = _mm256_extractf128_si256(sy, 1);
+    // Perform multiplication and create 16-bit values
+    const __m128i dotl = _mm_maddubs_epi16(axl, syl);
+    const __m128i doth = _mm_maddubs_epi16(axh, syh);
+    return sum_i16_pairs_float(doth, dotl);
+}
+
 // multiply int8_t, add results pairwise twice and return as float vector
 static inline __m256 mul_sum_i8_pairs_float(const __m256i x, const __m256i y) {
     const __m128i xl = _mm256_castsi256_si128(x);
@@ -2465,7 +2486,7 @@ static void ggml_vec_dot_q4_1_q8_1(const int n, float * restrict s, const void *
         const __m256i bx = bytes_from_nibbles_32(x[i].qs);
         const __m256i by = _mm256_loadu_si256( (const __m256i *)y[i].qs );
 
-        const __m256 xy = mul_sum_i8_pairs_float(bx, by);
+        const __m256 xy = mul_sum_us8_pairs_float(bx, by);
 
         // Accumulate d0*d1*x*y
 #if defined(__AVX2__)
@@ -2937,7 +2958,7 @@ static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void *
         const __m256 dy = _mm256_broadcast_ss(&y[i].d);
         const __m256i by = _mm256_loadu_si256((const __m256i *)y[i].qs);
 
-        const __m256 q = mul_sum_i8_pairs_float(bx, by);
+        const __m256 q = mul_sum_us8_pairs_float(bx, by);
 
         acc = _mm256_fmadd_ps(q, _mm256_mul_ps(dx, dy), acc);
     }
@@ -2971,7 +2992,7 @@ static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void *
         const __m256 dy = _mm256_broadcast_ss(&y[i].d);
         const __m256i by = _mm256_loadu_si256((const __m256i *)y[i].qs);
 
-        const __m256 q = mul_sum_i8_pairs_float(bx, by);
+        const __m256 q = mul_sum_us8_pairs_float(bx, by);
 
         acc = _mm256_add_ps(_mm256_mul_ps(q, _mm256_mul_ps(dx, dy)), acc);
     }