@@ -15441,7 +15441,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
1544115441        if (ftype == LLAMA_FTYPE_CQS && qs.params->attn_q_type < GGML_TYPE_COUNT) {
1544215442            new_type = qs.params->attn_q_type;
1544315443        }
15444-         if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) {
15444+         else  if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) {
1544515445            new_type = GGML_TYPE_IQ3_XXS;
1544615446        }
1544715447        else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
@@ -15453,7 +15453,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
1545315453        if (ftype == LLAMA_FTYPE_CQS && qs.params->ffn_down_type < GGML_TYPE_COUNT) {
1545415454            new_type = qs.params->ffn_down_type;
1545515455        }
15456-         else if       (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
15456+         else if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
1545715457        else if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S) {
1545815458            if (i_layer < n_layer/8) new_type = GGML_TYPE_Q4_K;
1545915459        }
@@ -15522,7 +15522,7 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n
1552215522        if (ftype == LLAMA_FTYPE_CQS && qs.params->attn_qkv_type < GGML_TYPE_COUNT) {
1552315523            new_type = qs.params->attn_qkv_type;
1552415524        }
15525-         if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L || ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) {
15525+         else  if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L || ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) {
1552615526            new_type = GGML_TYPE_Q4_K;
1552715527        }
1552815528        else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M) new_type = GGML_TYPE_Q5_K;
0 commit comments