[oceanio]

Problem and result

On a Mac, ./setup.sh sets up the original Qwen3.8-Flash-Next, Unsloth's UD files and agentionai's Gyro files only;
docs/MAC.md says "The Swift and Coder releases are not set up on Mac yet". Swift 1.5 itself runs on the Metal engine,
but setup cannot prepare it: its two shards cut through a layer (IQ3_S: layer 14's gate in shard 1, its up and down in shard 2), shard 1 also holds
the n-gram table, and the Metal engine maps one file.

With this PR, ./setup.sh --family swift downloads Swift 1.5 IQ3_S, repacks its shards into one GGUF with the existing
metal/tools/metal_repack.py (as setup already does for Unsloth's files), builds the pack and writes a config. Without
--family swift nothing changes: the size menu, the original model, Unsloth's and Gyro's files.

Related discussion

#2898 (feature request). No maintainer reply on scope yet: opened as a draft so the change can be read beside the request.

Approach

• Whether a model is repacked now follows repacks() (shards that cut through layers: Unsloth's 4-6, Swift's 2)
  instead of the shard count. The n-gram table stays in Swift's shard 1.
• --family takes swift beside qwen, unsloth and gyro. --family swift takes IQ3_S only, the size measured
  on a Mac; another size stops setup with the command to run.
  Images follow the original model's choice (on by default, --vision no turns them off), with Swift's own projector
  as on a PC: setup.py's Swift entry names UkisAI's mmproj-Swift-Qwen3.8-Flash-Next-BF16.gguf, and its size and
  SHA-256 (setup.py's pin) are added to VISION_SHA256.
• Swift's pack size lives in strata_control/mac_fit.py (MAC_SWIFT, mac_model(), and an optional family
  argument on model_need_gb / streams / model_note). The existing callers (setup.py's Control mode,
  tools/strata_mcp.py, strata launch) pass (model, ram) and see no change.
• No engine change: the pack, the MTP draft layer and the engine are the same. iq_pack writes Swift's 96 F32 routers
  as BF16 (all exact), and the draft layer is the original model's, as on a PC.
• docs/MAC.md: the --family swift option, what is measured, and the Families line.

Validation

Re-run on v0.2.6 (2026-10-26). Rebased onto main 4b9e05a2 (v0.2.6). Only docs/MAC.md conflicted (upstream rewrote it); the setup commits are the same lines.

• tools/test_setup_mac.py passes.
• Swift 1.5 IQ3_S runs on v0.2.6's own engine (0.1.50-metal; this PR changes no engine code). Needles: 9 of 9 found at 2K, 16K and 32K, depths 10/50/90%.
• With UkisAI's projector all three test pictures are read right: the shape counts, the invoice text word for word, and the bar order.

The rest of this section is the run on v0.2.2.

Environment: M5 Max 128 GB, macOS 26.7.1, Command Line Tools 27.0 (Apple clang 21).

Passed:

• python3 tools/test_setup_mac.py on 8943a2b (= main 011c9e6, 0.2.2, + this PR's 2 commits): 32 tests OK, 6 of
  them new
  (the original's and Unsloth's file names and repack decision unchanged; Swift's file names, its repack, its engine
  arguments; IQ3_S with --vision yes and no warning, its projector's pin equal to setup.py's; an unmeasured size
  stopping; the choices without --family swift).
• python3 -m unittest tools.test_strata_mcp.MacCatalog tools.test_strata_mcp.Helpers (the tests that import
  mac_fit): OK on this PR; python3 tools/test_setup_menu.py: OK.
• Swift 1.5 IQ3_S on the Metal engine (SC117's abliterated build, the same two-file layout as UkisAI's release),
  repacked file 54.8 GB: decode 93.1-110.4 tok/s on 4 prompts (Qwen IQ3_S 94.3-109.2 on the same Mac), prompt
  1,636-1,731 tok/s, 6.8K-token needles 3 of 3, 3 reasoning questions right with thinking on. Measured on the 0.1.45
  engine with the Mac build fix that 0.2.0 carries.
• The engine built from this branch (= main 011c9e6, 0.2.2, no engine change; cmake --build build-metal --target strata,
  0 errors) with the repacked Swift file, through this branch's serve/server.py at --max-context 262144: answers a
  short question ("Paris"), and tools/needle_bench.py --lengths 2k,16k,32k --depths 10,50,90 finds 9 of 9
  (1,644 / 15,070 / 31,643 prompt tokens). The same checks passed on 0.2.1.4 (main 9de04cd) before the rebase.
• Images, with the same engine and server and UkisAI's projector (SHA-256 equal to the pin), greedy:
  three test pictures read right (6 shapes counted by color, an invoice transcribed exactly, 5 bars ranked).

Failed: none observed.

Not run:

• tools/test_strata_mcp.py in full: its fake server uses port 8080, which a running Strata held on this Mac.
• A fresh end-to-end ./setup.sh --family swift on 0.2.1.4 (55 GB download and repack). It ran end to end
  (--gguf-dir) before the rebase, on 0.1.45 with the Mac build fix.
• UkisAI's own model files (SC117's build has the same layout; UkisAI's projector is the one checked above), other Macs, streaming on a smaller Mac, quality beyond 32K
  tokens, the KL harness, the Coder release.

Limits

• IQ3_S only; other Swift sizes stop setup with a message until they are measured on a Mac.
• Opt-in only: the size menu, the app and strata launch do not offer Swift.

🤖 Generated with Claude Code

[Niko1221]

Thank you for the Mac setup work. Mac support is part of the Mac plan, which is planned for a later release, with no date yet. This will stay open until then.