remove mistral arch

Author: juliendenize

gguf-py/gguf/constants.py

@@ -686,7 +686,6 @@ class MODEL_TENSOR(IntEnum):
     MODEL_ARCH.SMOLLM3:          "smollm3",
     MODEL_ARCH.LFM2:             "lfm2",
     MODEL_ARCH.DREAM:            "dream",
-    MODEL_ARCH.MISTRAL:          "mistral",
 }
 
 VISION_PROJECTOR_TYPE_NAMES: dict[VISION_PROJECTOR_TYPE, str] = {
@@ -2436,21 +2435,6 @@ class MODEL_TENSOR(IntEnum):
         MODEL_TENSOR.ATTN_V,
         MODEL_TENSOR.ATTN_OUT,
     ],
-    MODEL_ARCH.MISTRAL: [
-        MODEL_TENSOR.TOKEN_EMBD,
-        MODEL_TENSOR.OUTPUT_NORM,
-        MODEL_TENSOR.OUTPUT,
-        MODEL_TENSOR.ATTN_NORM,
-        MODEL_TENSOR.ATTN_Q,
-        MODEL_TENSOR.ATTN_K,
-        MODEL_TENSOR.ATTN_V,
-        MODEL_TENSOR.ATTN_OUT,
-        MODEL_TENSOR.FFN_GATE_INP,
-        MODEL_TENSOR.FFN_NORM,
-        MODEL_TENSOR.FFN_GATE,
-        MODEL_TENSOR.FFN_DOWN,
-        MODEL_TENSOR.FFN_UP,
-    ]
     # TODO
 }
 

src/llama-arch.cpp

@@ -86,7 +86,6 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
     { LLM_ARCH_SMOLLM3,          "smollm3"          },
     { LLM_ARCH_LFM2,             "lfm2"             },
     { LLM_ARCH_DREAM,            "dream"            },
-    { LLM_ARCH_MISTRAL,          "mistral"          },
     { LLM_ARCH_UNKNOWN,          "(unknown)"        },
 };
 
@@ -1887,24 +1886,6 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
             { LLM_TENSOR_TOKEN_EMBD_NORM,   "token_embd_norm" },
         }
     },
-    {
-        LLM_ARCH_MISTRAL,
-        {
-            { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
-            { LLM_TENSOR_OUTPUT_NORM,     "output_norm" },
-            { LLM_TENSOR_OUTPUT,          "output" },
-            { 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_GATE_INP,    "blk.%d.ffn_gate_inp" },
-            { LLM_TENSOR_FFN_NORM,        "blk.%d.ffn_norm" },
-            { LLM_TENSOR_FFN_GATE,        "blk.%d.ffn_gate" },
-            { LLM_TENSOR_FFN_DOWN,        "blk.%d.ffn_down" },
-            { LLM_TENSOR_FFN_UP,          "blk.%d.ffn_up" },
-        },
-    },
     {
         LLM_ARCH_UNKNOWN,
         {

src/llama-arch.h

@@ -90,7 +90,6 @@ enum llm_arch {
     LLM_ARCH_SMOLLM3,
     LLM_ARCH_LFM2,
     LLM_ARCH_DREAM,
-    LLM_ARCH_MISTRAL,
     LLM_ARCH_UNKNOWN,
 };
 

src/llama-model.cpp

@@ -570,8 +570,6 @@ void llama_model::load_hparams(llama_model_loader & ml) {
     // arch-specific KVs
     switch (arch) {
         case LLM_ARCH_LLAMA:
-        case LLM_ARCH_MISTRAL:
-            {
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
 
                 if (hparams.n_expert == 8) {
@@ -2019,7 +2017,6 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
             case LLM_ARCH_MINICPM:
             case LLM_ARCH_GRANITE:
             case LLM_ARCH_GRANITE_MOE:
-            case LLM_ARCH_MISTRAL:
                 {
                     tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
 
@@ -16738,7 +16735,6 @@ llm_graph_result_ptr llama_model::build_graph(
 
     switch (arch) {
         case LLM_ARCH_LLAMA:
-        case LLM_ARCH_MISTRAL:
             {
                 llm = std::make_unique<llm_build_llama>(*this, params, gf);
             } break;
@@ -17209,7 +17205,6 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
         case LLM_ARCH_SMOLLM3:
         case LLM_ARCH_ARCEE:
         case LLM_ARCH_ERNIE4_5:
-        case LLM_ARCH_MISTRAL:
             return LLAMA_ROPE_TYPE_NORM;
 
         // the pairs of head values are offset by n_rot/2