From 4c70a9ed7252e648befd750cfc185e41067b626f Mon Sep 17 00:00:00 2001
From: Daniel Han <danielhanchen@gmail.com>
Date: Thu, 7 May 2026 05:58:58 +0000
Subject: [PATCH] ci(mac): switch Mac GGUF jobs to UD-Q4_K_XL + bump UI turn
 timeout

The IQ3_XXS quants the Linux smoke uses are pathological at
temperature=0 on Apple Silicon Metal:

  - Qwen3.5-2B IQ3_XXS emits 'The The The...' for tool-call prompts
    (no tool_calls in the response, hits max_tokens).
  - gemma-4-E2B IQ3_XXS emits '<unused5><unused5>...' for any prompt
    (model degenerates to padding tokens).

Both are inference-path-correct but quant-degenerate; the Linux CPU
backend hides the issue. Bump both to UD-Q4_K_XL, the smallest
published variant that generates real text + well-formed tool calls
on M1. Inference time goes up modestly (CI is cache-warm so download
cost is one-shot per HF release).

Also bump STUDIO_UI_TURN_TIMEOUT_MS to 540s for the Mac UI job:
the macos-14 free runner is 3-5x slower than ubuntu-latest at
gemma-3-270m CPU inference, and the existing 180s ceiling crowded
turn 4 ('say tree').
---
 .../workflows/studio-mac-inference-smoke.yml   | 18 ++++++++++++++----
 .github/workflows/studio-mac-ui-smoke.yml      |  6 ++++++
 2 files changed, 20 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/studio-mac-inference-smoke.yml b/.github/workflows/studio-mac-inference-smoke.yml
index 0774ccd98d..b42e37ecc2 100644
--- a/.github/workflows/studio-mac-inference-smoke.yml
+++ b/.github/workflows/studio-mac-inference-smoke.yml
@@ -292,8 +292,13 @@ jobs:
     timeout-minutes: 25
     env:
       GGUF_REPO: unsloth/Qwen3.5-2B-GGUF
-      GGUF_VARIANT: UD-IQ3_XXS
-      GGUF_FILE: Qwen3.5-2B-UD-IQ3_XXS.gguf
+      # The Linux smoke uses UD-IQ3_XXS, but on Mac Metal that quant
+      # collapses for tool-call grammar at temperature=0 (model emits
+      # repeated padding tokens until max_tokens). UD-Q4_K_XL is the
+      # smallest published variant that produces well-formed tool
+      # calls + non-pathological text on M1.
+      GGUF_VARIANT: UD-Q4_K_XL
+      GGUF_FILE: Qwen3.5-2B-UD-Q4_K_XL.gguf
       STUDIO_PORT: '18898'
       HF_HOME: ${{ github.workspace }}/hf-cache
     steps:
@@ -607,8 +612,13 @@ jobs:
     timeout-minutes: 30
     env:
       GGUF_REPO: unsloth/gemma-4-E2B-it-GGUF
-      GGUF_VARIANT: UD-IQ3_XXS
-      GGUF_FILE: gemma-4-E2B-it-UD-IQ3_XXS.gguf
+      # Linux smoke uses UD-IQ3_XXS, but on Mac Metal that gemma-4
+      # quant emits sentinel tokens (<unused5>) for any prompt at
+      # temperature=0 -- inference path is fine, the quant itself is
+      # broken on Metal. UD-Q4_K_XL is the smallest published variant
+      # that generates real text on M1.
+      GGUF_VARIANT: UD-Q4_K_XL
+      GGUF_FILE: gemma-4-E2B-it-UD-Q4_K_XL.gguf
       MMPROJ_FILE: mmproj-F16.gguf
       STUDIO_PORT: '18899'
       HF_HOME: ${{ github.workspace }}/hf-cache
diff --git a/.github/workflows/studio-mac-ui-smoke.yml b/.github/workflows/studio-mac-ui-smoke.yml
index 71a906f996..df89970540 100644
--- a/.github/workflows/studio-mac-ui-smoke.yml
+++ b/.github/workflows/studio-mac-ui-smoke.yml
@@ -130,6 +130,10 @@ jobs:
           BASE_URL: http://127.0.0.1:18896
           PW_ART_DIR: logs/playwright
           STUDIO_UI_STRICT: '1'
+          # macos-14 free runner is 3 vCPU / 7 GB / no Metal-accel
+          # available to llama.cpp from CI; gemma-3-270m turn latency
+          # has been observed to crowd the 180s default. Triple it.
+          STUDIO_UI_TURN_TIMEOUT_MS: '540000'
         run: |
           mkdir -p logs/playwright
           python tests/studio/playwright_chat_ui.py
@@ -174,6 +178,8 @@ jobs:
           STUDIO_NEW_PW: ${{ env.STUDIO_EXTRA_NEW_PW }}
           PW_ART_DIR: logs/playwright_extra
           STUDIO_UI_STRICT: '1'
+          # See "Drive the chat UI" step.
+          STUDIO_UI_TURN_TIMEOUT_MS: '540000'
           GGUF_REPO: ${{ env.GGUF_REPO }}
           GGUF_VARIANT: ${{ env.GGUF_VARIANT }}
         run: |