fix name

* jinja : parse unary +/- before variables

Lexer already emits unary_operator for -n / +n, and runtime executes
unary -. Parse them at multiplicative precedence so slices like
items[:-n] and GigaChat indent[:-indent_factor] work.

* jinja : keep filters/tests outside unary operands

Unary +/- must bind only the primary/postfix operand so -n|abs is
(-n)|abs, not -(n|abs). Add unary + and filter/test regression coverage.

Signed-off-by: sinksilk <[email protected]>

---------

Signed-off-by: sinksilk <[email protected]>

.github/workflows/ci-self-hosted-cuda.yml

@@ -88,8 +88,8 @@ jobs:
           hf_bucket: ggml-org/cache
           save: true
 
-  gpu-rocm:
-    runs-on: [self-hosted, Linux, AMD]
+  gpu-rocm-gfx1151:
+    runs-on: [self-hosted, Linux, gfx1151]
 
     steps:
       - name: Clone
@@ -108,17 +108,47 @@ jobs:
           rocminfo
           GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS=gfx1151 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
 
-  # TODO: provision AMD GPU machine
-  # amd-rocm:
-  #   runs-on: [self-hosted, Linux, AMD]
+  gpu-rocm-gfx1201:
+    runs-on: [self-hosted, Linux, gfx1201]
+    container: "rocm/dev-ubuntu-24.04:7.2.4-complete"
 
-  #   steps:
-  #     - name: Clone
-  #       id: checkout
-  #       uses: actions/checkout@v6
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
 
-  #     - name: Test
-  #       id: ggml-ci
-  #       run: |
-  #         amd-smi static
-  #         GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+      - name: Install dependencies
+        run: |
+          apt update
+          apt install -y build-essential git git-lfs jq cmake libssl-dev time unzip wget python3 python3-venv python3-pip
+
+      - name: ccache
+        uses: ggml-org/[email protected]
+        with:
+          restore: false
+          save: false
+
+      - name: ccache-buckets-restore
+        uses: ./.github/actions/ccache-buckets
+        with:
+          key: self-hosted-gpu-rocm-gfx1201
+          folder: llama.cpp
+          hf_bucket: ggml-org/cache
+
+      - name: Test
+        id: ggml-ci
+        run: |
+          rocminfo
+          GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS=gfx1201 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+
+      - name: ccache-buckets-save
+        if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+        uses: ./.github/actions/ccache-buckets
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN_CACHE_OUTPUT }}
+        with:
+          key: self-hosted-gpu-rocm-gfx1201
+          folder: llama.cpp
+          evict-old-files: 1d
+          hf_bucket: ggml-org/cache
+          save: true

.github/workflows/ci-self-hosted-vulkan.yml

@@ -185,17 +185,47 @@ jobs:
           # a valid python environment for testing
           LLAMA_FATAL_WARNINGS=OFF GG_BUILD_NINJA=1 GG_BUILD_VULKAN=1 GG_BUILD_LOW_PERF=1 ./ci/run.sh ./results/llama.cpp ./mnt/llama.cpp
 
-  # TODO: provision AMD GPU machine
-  # amd-vulkan:
-  #   runs-on: [self-hosted, Linux, AMD]
+  gpu-vulkan-amd:
+    runs-on: [self-hosted, Linux, gfx1201]
+    container: "ubuntu:26.04"
 
-  #   steps:
-  #     - name: Clone
-  #       id: checkout
-  #       uses: actions/checkout@v6
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
 
-  #     - name: Test
-  #       id: ggml-ci
-  #       run: |
-  #         vulkaninfo --summary
-  #         GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+      - name: Install dependencies
+        run: |
+          apt update
+          apt install -y build-essential git git-lfs jq cmake libxcb-xinput0 libxcb-xinerama0 libxcb-cursor-dev libvulkan-dev glslc spirv-headers vulkan-tools mesa-vulkan-drivers libglvnd0 libgl1 libglx0 libegl1 libgles2 libssl-dev time unzip wget python3 python3-venv python3-pip
+
+      - name: ccache
+        uses: ggml-org/[email protected]
+        with:
+          restore: false
+          save: false
+
+      - name: ccache-buckets-restore
+        uses: ./.github/actions/ccache-buckets
+        with:
+          key: self-hosted-vulkan-amd
+          folder: llama.cpp
+          hf_bucket: ggml-org/cache
+
+      - name: Test
+        id: ggml-ci
+        run: |
+          vulkaninfo --summary
+          GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp
+
+      - name: ccache-buckets-save
+        if: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+        uses: ./.github/actions/ccache-buckets
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN_CACHE_OUTPUT }}
+        with:
+          key: self-hosted-vulkan-amd
+          folder: llama.cpp
+          evict-old-files: 1d
+          hf_bucket: ggml-org/cache
+          save: true

common/jinja/parser.cpp

@@ -437,7 +437,8 @@ private:
     }
 
     statement_ptr parse_filter_expression() {
-        auto operand = parse_call_member_expression();
+        // Filters/tests bind outside unary so -n|abs is (-n)|abs, not -(n|abs).
+        auto operand = parse_unary_expression();
         while (is(token::pipe)) {
             size_t start_pos = current;
             ++current; // consume pipe
@@ -448,6 +449,15 @@ private:
         return operand;
     }
 
+    statement_ptr parse_unary_expression() {
+        if (is(token::unary_operator)) {
+            size_t start_pos = current;
+            auto op = next();
+            return mk_stmt<unary_expression>(start_pos, op, parse_unary_expression());
+        }
+        return parse_call_member_expression();
+    }
+
     statement_ptr parse_call_member_expression() {
         // Handle member expressions recursively
         auto member = parse_member_expression(parse_primary_expression());

common/jinja/runtime.cpp

@@ -450,6 +450,11 @@ value unary_expression::execute_impl(context & ctx) const {
         } else {
             throw std::runtime_error("Unary - operator requires numeric operand");
         }
+    } else if (op.value == "+") {
+        if (is_val<value_int>(operand_val) || is_val<value_float>(operand_val)) {
+            return operand_val;
+        }
+        throw std::runtime_error("Unary + operator requires numeric operand");
     }
 
     throw std::runtime_error("Unknown unary operator '" + op.value + "'");

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -3584,7 +3584,6 @@ int ggml_metal_op_flash_attn_ext(ggml_metal_op_t ctx, int idx) {
         auto cfg = use_sparse
                 ? ggml_metal_tuning::fa_vec_baseline_cfg((int) ne00, (int) ne20)
                 : ggml_metal_tuning::fa_vec_pick(
-                          props_dev->device_id,
                           props_dev->gpu_family,
                           (int) op->src[1]->type,
                           (int) ne00, (int) ne20,   // dk, dv (ne00 == dk for FA)

ggml/src/ggml-metal/ggml-metal-tuning.h

@@ -1,6 +1,5 @@
 #pragma once
 
-#include "ggml-metal-device.h"  // enum ggml_metal_device_id
 #include "ggml.h"
 
 #include <cstdint>
@@ -32,7 +31,7 @@ constexpr int8_t FA_VEC_DOMAIN_DECODE = 0;  // ne01 == 1
 constexpr int8_t FA_VEC_DOMAIN_BATCH  = 1;  // ne01 >= 2
 
 struct fa_vec_key_t {
-    int8_t  device_id;
+    int8_t  family;
     int8_t  dtype;
     int16_t dk;
     int16_t dv;
@@ -70,8 +69,7 @@ void         fa_vec_set_override(fa_vec_cfg_t cfg);
 void         fa_vec_clear_override();
 fa_vec_cfg_t fa_vec_baseline_cfg(int dk, int dv);
 
-// device_id selects a per-SKU row; on a miss, gpu_family (0 if unknown) maps to a representative
-// SKU and the table is retried. No match -> baseline.
-fa_vec_cfg_t fa_vec_pick(enum ggml_metal_device_id device_id, int gpu_family, int dtype, int dk, int dv, int64_t ne11, int64_t ne01);
+// Keyed by Apple GPU family; an untuned family matches no row and gets the baseline.
+fa_vec_cfg_t fa_vec_pick(int gpu_family, int dtype, int dk, int dv, int64_t ne11, int64_t ne01);
 
 }  // namespace ggml_metal_tuning

tests/test-jinja.cpp

@@ -458,6 +458,49 @@ static void test_expressions(testing & t) {
         "['b']"
     );
 
+    test_template(t, "array slice negative variable",
+        "{{ items[:-n]|string }}",
+        {{"items", json::array({"a", "b", "c"})}, {"n", 1}},
+        "['a', 'b']"
+    );
+
+    test_template(t, "array slice negative variable indent",
+        "{{ indent[:-indent_factor] }}",
+        {{"indent", "    "}, {"indent_factor", 2}},
+        "  "
+    );
+
+    test_template(t, "unary minus variable",
+        "{{ -n }}",
+        {{"n", 3}},
+        "-3"
+    );
+
+    test_template(t, "unary plus variable",
+        "{{ +n }}",
+        {{"n", -3}},
+        "-3"
+    );
+
+    test_template(t, "unary plus float",
+        "{{ +x }}",
+        {{"x", -1.5}},
+        "-1.5"
+    );
+
+    // Unary binds tighter than filter: -n|abs == (-n)|abs, not -(n|abs)
+    test_template(t, "unary minus then abs filter",
+        "{{ -n|abs }}",
+        {{"n", -3}},
+        "3"
+    );
+
+    test_template(t, "unary minus then number test",
+        "{{ -n is number }}",
+        {{"n", 3}},
+        "True"
+    );
+
     test_template(t, "array slice step",
         "{{ items[::2]|string }}",
         {{"items", json::array({"a", "b", "c"})}},

tools/tuning/README.md

@@ -26,10 +26,13 @@ Sweep the grid (6 dtypes x 10 head sizes x 4 KV depths x 9 batch widths; a few h
 ./build/bin/ggml-metal-tuning fa-vec > fa_vec_rows.txt 2> fa_vec_sweep.log
 ```
 
-`fa_vec_rows.txt` holds nothing but table rows, ready to paste into `fa_vec_tuned_table`: the min-max-regret target, the aggregate benefit gate, the short-KV drop and the pointwise compression are already applied.
+`fa_vec_rows.txt` holds nothing but table rows: the min-max-regret target, the aggregate benefit gate, the short-KV drop and the pointwise compression are already applied.
+The rows carry the SKU token the runtime reported, but `fa_vec_tuned_table` is keyed by Apple GPU family, so that column has to be retagged before the rows compile.
+Your family number is on the `MTLGPUFamilyApple<N>` line the backend logs at init, near the top of `fa_vec_sweep.log`; `N` is the value, and the `MTLGPUFamilyCommon`/`MTLGPUFamilyMetal` lines beside it are not it.
+If your log is the only sweep for that family, its rows become the family's segment; where the family already has rows, post the log and let the two be compared before anything is replaced.
 A config represents a bucket only if it is no slower than the baseline config at every point that bucket covers, so a config that wins on average but loses at one batch width leaves its bucket at baseline.
 `fa_vec_sweep.log` holds the per-cell timings, bucket coverage, noise floor, any cooldown activity, and every config the no-harm rule refused together with the point that refused it.
-Post both: the log is what makes the rows reviewable.
+Post both, always: the rows now speak for every device in the family, so the log is what makes them reviewable.
 
 Long sweeps can be split.
 `--dtype f16,q4_0` and `--dk 128,192` restrict the grid, and the emitted rows for one `(dtype, head size)` do not depend on the others.