Function ggml_soft_max_impl

subprojects/llama.cpp/ggml/src/ggml.c:3935–3967

// ggml_soft_max

static struct ggml_tensor * ggml_soft_max_impl(
        struct ggml_context * ctx,
        struct ggml_tensor  * a,
        struct ggml_tensor  * mask,
        float                 scale,
        float                 max_bias,
        bool                  inplace) {
    GGML_ASSERT(ggml_is_contiguous(a));

    if (mask) {
        GGML_ASSERT(mask->type == GGML_TYPE_F16 || mask->type == GGML_TYPE_F32);
        GGML_ASSERT(ggml_is_contiguous(mask));
        GGML_ASSERT(mask->ne[0] == a->ne[0]);
        GGML_ASSERT(mask->ne[1] >= a->ne[1]);
        GGML_ASSERT(a->ne[2]%mask->ne[2] == 0);
        GGML_ASSERT(a->ne[3]%mask->ne[3] == 0);
    }

    if (max_bias > 0.0f) {
        GGML_ASSERT(mask);
    }

    struct ggml_tensor * result = inplace ? ggml_view_tensor(ctx, a) : ggml_dup_tensor(ctx, a);

    float params[] = { scale, max_bias };
    ggml_set_op_params(result, params, sizeof(params));

    result->op     = GGML_OP_SOFT_MAX;
    result->src[0] = a;
    result->src[1] = mask;

    return result;
}

struct ggml_tensor * ggml_soft_max(
        struct ggml_context * ctx,