fix: revert some changes

Author: Joan Martinez

convert-hf-to-gguf.py

@@ -92,7 +92,6 @@ def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
 
     def set_gguf_parameters(self):
         self.gguf_writer.add_name(self.dir_model.name)
-        print(f'self.block_count {self.block_count}')
         self.gguf_writer.add_block_count(self.block_count)
 
         if (n_ctx := self.find_hparam(["max_position_embeddings", "n_ctx"], optional=True)) is not None:
@@ -138,7 +137,6 @@ def set_gguf_parameters(self):
     def write_tensors(self):
         block_count = self.hparams.get("n_layers", self.hparams.get("num_hidden_layers", self.hparams.get("n_layer")))
         tensor_map = gguf.get_tensor_name_map(self.model_arch, block_count)
-        print(f'Block_count {block_count} with tensor_map {tensor_map}')
         for name, data_torch in self.get_tensors():
             # we don't need these
             if name.endswith((".attention.masked_bias", ".attention.bias", ".attention.rotary_emb.inv_freq")):
@@ -2188,6 +2186,9 @@ def get_tensors(self):
             yield name, data
 
 
+JinaBertForMaskedML = JinaBertModel
+
+
 @Model.register("GemmaForCausalLM")
 class GemmaModel(Model):
     model_arch = gguf.MODEL_ARCH.GEMMA

ggml.c

@@ -12476,8 +12476,9 @@ static void ggml_compute_forward_alibi_f32(
 
         for (int64_t i = 0; i < ne0; i++) {
             for (int64_t j = 0; j < ne1; j++) {
+                float * const src = (float *)((char *) src0->data + i*nb0 + j*nb1 + k*nb2);
                 float *      pdst = (float *)((char *)  dst->data + i*nb0 + j*nb1 + k*nb2);
-                pdst[0] = -1.0f * i * m_k;
+                pdst[0] = i * m_k + src[0];
             }
         }
     }