Move quantization to new backend (#10363)

* Move quantization logic to GGML via new backend

This moves the model aware logic to Go code and calls GGMLs quantization code for model creation.

* Remove "add model quantizations"

This is no longer needed now that quantization is implemented in Go+GGML code directly.

llama/patches/0016-add-model-quantizations.patch

@@ -1,96 +0,0 @@
-From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
-From: jmorganca <jmorganca@gmail.com>
-Date: Tue, 8 Apr 2025 20:39:32 -0700
-Subject: [PATCH] add model quantizations
-
-a temporary patch to add model quantization for
-models not supported in llama.cpp
----
- src/llama-arch.cpp  | 17 +++++++++++++++++
- src/llama-arch.h    |  1 +
- src/llama-model.cpp |  2 ++
- src/llama-quant.cpp |  4 ++++
- 4 files changed, 24 insertions(+)
-
-diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp
-index eb7b5325..df42d1a5 100644
---- a/src/llama-arch.cpp
-+++ b/src/llama-arch.cpp
-@@ -74,6 +74,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
-     { LLM_ARCH_WAVTOKENIZER_DEC, "wavtokenizer-dec" },
-     { LLM_ARCH_PLM,              "plm"              },
-     { LLM_ARCH_BAILINGMOE,       "bailingmoe"       },
-+    { LLM_ARCH_MISTRAL3,         "mistral3"         },
-     { LLM_ARCH_UNKNOWN,          "(unknown)"        },
- };
- 
-@@ -1606,6 +1607,22 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
-             { LLM_TENSOR_FFN_UP_SHEXP,       "blk.%d.ffn_up_shexp" },
-         },
-     },
-+    {
-+        LLM_ARCH_MISTRAL3,
-+        {
-+            { LLM_TENSOR_TOKEN_EMBD,  "token_embd" },
-+            { LLM_TENSOR_OUTPUT_NORM, "output_norm" },
-+            { LLM_TENSOR_ATTN_NORM,   "blk.%d.attn_norm" },
-+            { LLM_TENSOR_ATTN_Q,      "blk.%d.attn_q" },
-+            { LLM_TENSOR_ATTN_K,      "blk.%d.attn_k" },
-+            { LLM_TENSOR_ATTN_V,      "blk.%d.attn_v" },
-+            { LLM_TENSOR_ATTN_OUT,    "blk.%d.attn_output" },
-+            { LLM_TENSOR_FFN_NORM,    "blk.%d.ffn_norm" },
-+            { LLM_TENSOR_FFN_GATE,    "blk.%d.ffn_gate" },
-+            { LLM_TENSOR_FFN_UP,      "blk.%d.ffn_up" },
-+            { LLM_TENSOR_FFN_DOWN,    "blk.%d.ffn_down" },
-+        }
-+    },
-     {
-         LLM_ARCH_UNKNOWN,
-         {
-diff --git a/src/llama-arch.h b/src/llama-arch.h
-index bc8a4f0b..bda9d071 100644
---- a/src/llama-arch.h
-+++ b/src/llama-arch.h
-@@ -76,6 +76,7 @@ enum llm_arch {
-     LLM_ARCH_CHAMELEON,
-     LLM_ARCH_SOLAR,
-     LLM_ARCH_WAVTOKENIZER_DEC,
-+    LLM_ARCH_MISTRAL3,
-     LLM_ARCH_PLM,
-     LLM_ARCH_BAILINGMOE,
-     LLM_ARCH_UNKNOWN,
-diff --git a/src/llama-model.cpp b/src/llama-model.cpp
-index 9d099f11..ef70486d 100644
---- a/src/llama-model.cpp
-+++ b/src/llama-model.cpp
-@@ -1437,6 +1437,7 @@ void llama_model::load_hparams(llama_model_loader & ml) {
-                     default: type = LLM_TYPE_UNKNOWN;
-                 }
-             } break;
-+        case LLM_ARCH_MISTRAL3: break;
-         default: throw std::runtime_error("unsupported model architecture");
-     }
- 
-@@ -13751,6 +13752,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
-         case LLM_ARCH_CHAMELEON:
-         case LLM_ARCH_SOLAR:
-         case LLM_ARCH_BAILINGMOE:
-+        case LLM_ARCH_MISTRAL3:
-             return LLAMA_ROPE_TYPE_NORM;
- 
-         // the pairs of head values are offset by n_rot/2
-diff --git a/src/llama-quant.cpp b/src/llama-quant.cpp
-index 223e1f3f..8ae6dde8 100644
---- a/src/llama-quant.cpp
-+++ b/src/llama-quant.cpp
-@@ -744,6 +744,10 @@ static void llama_model_quantize_impl(const std::string & fname_inp, const std::
-         // This used to be a regex, but <regex> has an extreme cost to compile times.
-         bool quantize = name.rfind("weight") == name.size() - 6; // ends with 'weight'?
- 
-+        // don't quantize vision stuff
-+        quantize &= name.find("v.") == std::string::npos;
-+        quantize &= name.find("mm.") == std::string::npos;
-+
-         // quantize only 2D and 3D tensors (experts)
-         quantize &= (ggml_n_dims(tensor) >= 2);
-