tuieval 0.2.0.dev62__tar.gz → 0.2.0.dev65__tar.gz

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/CHANGELOG.md

@@ -2,6 +2,7 @@
 
 ## Unreleased
 
+- Nvidia and AMD GPUs on Linux: VRAM from `nvidia-smi`, `amd-smi` or `rocm-smi` sizes models (`gpu_memory_gb` overrides it), runs record peak VRAM, and `tuieval doctor` shows the GPU.
 - Results has four tabs: PTA index, Production readiness, Per question and Failures. The PTA table adds tok/s, TTFT and memory; Per question can show only where models disagree; the settings warning is one line. Scorecard, pairwise, history and test quality stay in `tuieval compare`, `history` and `items`.
 - The run's ETA adds up each pack's answers left × that model's time per answer there (its history, then its own times once it has 5 answers), instead of extrapolating from the first few answers.
 - Faster start on Macs: the GPU core count comes from `ioreg` instead of `system_profiler`, which took seconds on some Macs.

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: tuieval
-Version: 0.2.0.dev62
+Version: 0.2.0.dev65
 Summary: Evaluate local and frontier LLMs on your own questions: accuracy, speed, tokens and PASS/FAIL verdicts, in the terminal.
 Project-URL: Homepage, https://github.com/ashe-wb/tuieval
 Project-URL: Issues, https://github.com/ashe-wb/tuieval/issues
@@ -51,7 +51,7 @@ Whatever runs your models, the first steps are short:
 | You have | Do this |
 |---|---|
 | **LM Studio, Ollama or vLLM** | Start its server and load a model, then `tuieval init my-evals`: it finds the model. Later: `tuieval add` lists new ones. |
-| **GGUF files** | Install llama.cpp (`brew install llama.cpp`), then `tuieval add ~/models/Some-Model-Q4_K_M.gguf`. tuieval starts `llama-server` for it, sized to your machine. |
+| **GGUF files** | Install llama.cpp (`brew install llama.cpp`, or a CUDA/ROCm build on Linux), then `tuieval add ~/models/Some-Model-Q4_K_M.gguf`. tuieval starts `llama-server` for it, sized to your Mac's memory or your Nvidia/AMD card's VRAM. |
 | **An OpenRouter key** | `export OPENROUTER_API_KEY=…`, then `tuieval run --tier smoke --only openrouter:<model id>`. |
 
 Stuck? `tuieval doctor` checks servers, models, packs and keys, and says what to fix.
@@ -76,7 +76,7 @@ Afterwards, every answer keeps its reasoning. Enter on a result opens it with th
 - **Verdicts you can act on.** A pack passes only with zero critical failures over enough trials, and an accuracy whose 95% lower bound clears your bar. *INCONCLUSIVE* says what evidence is missing.
 - **Two tiers:** *Screen* a sample of every pack to drop weak models fast, then *Certify* finalists on every test with repeats. Certification reuses the screening answers.
 - **Honest numbers:** every request is a fresh single-turn conversation, with prompt caching off and repeat rounds in different orders. Each answer records what was sent, and the run warns about reused prompts or identical repeats.
-- **Per-machine speed.** A fit check picks the largest context that fits your Mac's GPU memory, `tuieval tune` finds the fastest speed-only server flags (with a guard that rejects flags that change answers), and readiness includes a *Fast enough?* table per machine, measured or projected.
+- **Per-machine speed.** A fit check picks the largest context that fits your Mac's GPU memory or your Nvidia/AMD card's VRAM, `tuieval tune` finds the fastest speed-only server flags (with a guard that rejects flags that change answers), and readiness includes a *Fast enough?* table per machine, measured or projected.
 - **Tests that check themselves.** Every test carries a reference answer and known-wrong answers, and `tuieval selftest` checks the grader accepts the first and rejects the second, and that each gate is reachable at all.
 - **History.** Verdict changes are appended to `results/verdicts.jsonl`, and replaced results are moved to `history/`, never overwritten.
 

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/README.md

@@ -25,7 +25,7 @@ Whatever runs your models, the first steps are short:
 | You have | Do this |
 |---|---|
 | **LM Studio, Ollama or vLLM** | Start its server and load a model, then `tuieval init my-evals`: it finds the model. Later: `tuieval add` lists new ones. |
-| **GGUF files** | Install llama.cpp (`brew install llama.cpp`), then `tuieval add ~/models/Some-Model-Q4_K_M.gguf`. tuieval starts `llama-server` for it, sized to your machine. |
+| **GGUF files** | Install llama.cpp (`brew install llama.cpp`, or a CUDA/ROCm build on Linux), then `tuieval add ~/models/Some-Model-Q4_K_M.gguf`. tuieval starts `llama-server` for it, sized to your Mac's memory or your Nvidia/AMD card's VRAM. |
 | **An OpenRouter key** | `export OPENROUTER_API_KEY=…`, then `tuieval run --tier smoke --only openrouter:<model id>`. |
 
 Stuck? `tuieval doctor` checks servers, models, packs and keys, and says what to fix.
@@ -50,7 +50,7 @@ Afterwards, every answer keeps its reasoning. Enter on a result opens it with th
 - **Verdicts you can act on.** A pack passes only with zero critical failures over enough trials, and an accuracy whose 95% lower bound clears your bar. *INCONCLUSIVE* says what evidence is missing.
 - **Two tiers:** *Screen* a sample of every pack to drop weak models fast, then *Certify* finalists on every test with repeats. Certification reuses the screening answers.
 - **Honest numbers:** every request is a fresh single-turn conversation, with prompt caching off and repeat rounds in different orders. Each answer records what was sent, and the run warns about reused prompts or identical repeats.
-- **Per-machine speed.** A fit check picks the largest context that fits your Mac's GPU memory, `tuieval tune` finds the fastest speed-only server flags (with a guard that rejects flags that change answers), and readiness includes a *Fast enough?* table per machine, measured or projected.
+- **Per-machine speed.** A fit check picks the largest context that fits your Mac's GPU memory or your Nvidia/AMD card's VRAM, `tuieval tune` finds the fastest speed-only server flags (with a guard that rejects flags that change answers), and readiness includes a *Fast enough?* table per machine, measured or projected.
 - **Tests that check themselves.** Every test carries a reference answer and known-wrong answers, and `tuieval selftest` checks the grader accepts the first and rejects the second, and that each gate is reachable at all.
 - **History.** Verdict changes are appended to `results/verdicts.jsonl`, and replaced results are moved to `history/`, never overwritten.
 

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/docs/models.md

@@ -69,9 +69,21 @@ The best server flags differ per model and per machine, so tuieval splits them b
 
 ## Machines
 
-- **Machine id** is detected automatically (e.g. `m3max-64gb`, `m2-16gb`; set `EVALS_MACHINE` to rename it). Every result records the machine, the server version and the exact speed flags used. `tuieval machines` lists this machine and every other machine that has run the evals (they record themselves in `tuning/`).
-- **Fit check (Apple Silicon, llama):** before starting a GGUF, tuieval reads its header (layers, KV heads, hybrid attention layers) and picks the largest context that fits this machine's GPU memory, capped at `max_ctx`. A model that can't fit at 8k context is skipped with *doesn't fit on <machine>* instead of swapping. Packs that need more context than fits are skipped. It never quantizes the KV cache on its own, since that changes answers. Servers that size their own memory (their `cmd` doesn't take `{ctx}`) are left alone: their context is the model's `max_context`; `fit_check` on the server changes that.
-- **Per-machine settings** go under `[machines.<id>]`: `memory_headroom_gb` (GPU memory kept free for macOS, default 4), `gpu_residency_gb` (see the stall guard) and `parallel_models` (see below).
+- **Machine id** is detected automatically (e.g. `m3max-64gb`, `m2-16gb`, or `rtx4090-128gb` on a Linux machine with a discrete GPU; set `EVALS_MACHINE` to rename it). Every result records the machine, the server version and the exact speed flags used. `tuieval machines` lists this machine and every other machine that has run the evals (they record themselves in `tuning/`).
+- **Fit check (llama):** before starting a GGUF, tuieval reads its header (layers, KV heads, hybrid attention layers) and picks the largest context that fits this machine's GPU memory (a Mac's unified memory, or a discrete GPU's VRAM), capped at `max_ctx`. A model that can't fit at 8k context is skipped with *doesn't fit on <machine>* instead of swapping. Packs that need more context than fits are skipped. It never quantizes the KV cache on its own, since that changes answers. Servers that size their own memory (their `cmd` doesn't take `{ctx}`) are left alone: their context is the model's `max_context`; `fit_check` on the server changes that.
+- **Per-machine settings** go under `[machines.<id>]`: `memory_headroom_gb` (GPU memory kept free: default 4 on a Mac, 1.5 in a discrete GPU's VRAM), `gpu_memory_gb` (the GPU memory to size models against, when detection gets it wrong or you want to hold some back), `gpu_residency_gb` (see the stall guard) and `parallel_models` (see below).
+
+### Discrete GPUs (Linux: Nvidia, AMD)
+
+With `nvidia-smi` (Nvidia) or `amd-smi` / `rocm-smi` (AMD) installed, tuieval reads each card's name and VRAM. Several cards add up, since llama.cpp splits a model across them. `tuieval doctor` shows what it found.
+
+- Models are sized to fit **entirely in VRAM** (servers keep `-ngl 99`); splitting a model between GPU and CPU isn't supported. A model that doesn't fit is skipped with what it needs and what's free.
+- **Memory** in results is the server's peak VRAM: its own processes' on Nvidia, the whole machine's on AMD (its tools don't report it per process reliably, so it's only exact when nothing else uses the GPU).
+- `tuieval tune` rejects flags that leave VRAM nearly full. The macOS memory-pressure checks and the stall guard don't apply.
+- A machine that recorded itself before its GPU was detected keeps that id, so its results and tuning still match; `tuieval doctor` names the new id if you want to switch (`EVALS_MACHINE`).
+- Without the vendor tools, models are sized against system RAM, as on any other machine; set `gpu_memory_gb` instead.
+
+Windows isn't supported (WSL2 works like Linux).
 
 ## Several models at a time
 

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/_version.py

@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
 commit_id: str | None
 __commit_id__: str | None
 
-__version__ = version = '0.2.0.dev62'
-__version_tuple__ = version_tuple = (0, 2, 0, 'dev62')
+__version__ = version = '0.2.0.dev65'
+__version_tuple__ = version_tuple = (0, 2, 0, 'dev65')
 
 __commit_id__ = commit_id = None

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/compare.py

@@ -35,7 +35,8 @@ def _load_native(data):
     m, pack, run = data["model"], data["pack"], data["run"]
     info = {"label": m["label"], "served": m.get("served_name"), "tags": m.get("tags") or [],
             "settings": data.get("settings", {}), "load_s": run.get("load_s"),
-            "peak_rss_mb": run.get("peak_rss_mb"), "pack": pack["name"], "pack_fp": pack["fingerprint"],
+            "peak_rss_mb": run.get("peak_rss_mb"), "vram_peak_gb": run.get("vram_peak_gb"),
+            "pack": pack["name"], "pack_fp": pack["fingerprint"],
             # hosted (OpenRouter) models: the provider endpoint and its declared quantization
             "endpoint": run.get("endpoint"), "quantization": run.get("quantization"),
             "loaded_alongside": run.get("loaded_alongside") or []}

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/doctor.py

@@ -10,6 +10,7 @@ import platform
 import sys
 
 from . import engine
+from . import machines
 from . import workspace
 
 OK, WARN, BAD = "✓", "!", "✗"
@@ -44,6 +45,22 @@ def check(models_path=None):
         return r.bad
     r.line(OK, f"workspace {e.root}")
 
+    r.head("Machine")
+    mc = e.machine()
+    r.line(OK, f"{mc.id}: {mc.summary}")
+    if mc.discrete:
+        r.line(OK, f"models are sized to {mc.gpu_limit_gb:.0f} GB of VRAM, keeping {e.headroom_gb(mc):g} GB free "
+                   "(memory_headroom_gb, gpu_memory_gb under [machines.<id>] change it)")
+        new = machines.gpu_id(machines.discrete_gpus(), mc.ram_gb)
+        if mc.id != new and not os.environ.get("EVALS_MACHINE"):
+            r.line(WARN, f"keeping the id {mc.id} this machine had before its GPU was detected, so earlier results "
+                         "and tuning still match", f"to switch: EVALS_MACHINE={new} (results under the old id "
+                                                   "then count as another machine)")
+    elif platform.system() == "Linux" and any(os.path.exists(p) for p in ("/dev/nvidia0", "/dev/kfd")):
+        r.line(WARN, "this machine seems to have an Nvidia or AMD GPU, but nvidia-smi / amd-smi / rocm-smi "
+                     "isn't installed, so models are sized against system RAM",
+               "install the vendor's tools, or set gpu_memory_gb under [machines.<id>] in models.toml")
+
     r.head("Packs")
     for err in e.pack_errors:
         r.line(BAD, err, "fix the pack, then: tuieval selftest")

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/engine.py

@@ -1191,7 +1191,19 @@ class Engine:
 
     # ---- serving on this machine
     def machine(self):
-        return machines.detect()
+        """This machine, with its models.toml [machines.<id>] gpu_memory_gb applied. A Linux machine
+        that recorded itself before its GPU was detected keeps that id, so its results and tuning
+        still match (EVALS_MACHINE renames it)."""
+        mc = machines.detect()
+        if mc.discrete and not os.environ.get("EVALS_MACHINE"):
+            old = machines.machine_id(mc.chip, mc.ram_gb)
+            if os.path.isdir(os.path.join(self.tuning_dir, old)) and \
+                    not os.path.isdir(os.path.join(self.tuning_dir, mc.id)):
+                mc = dataclasses.replace(mc, id=old)
+        override = self.cfg.get("machines", {}).get(mc.id, {}).get("gpu_memory_gb")
+        if override:
+            mc = dataclasses.replace(mc, gpu_limit_gb=float(override))
+        return mc
 
     def machines(self):
         """This machine first, then every other machine recorded in tuning/ (synced)."""
@@ -1203,9 +1215,14 @@ class Engine:
         return [here] + [mc for mc in profiles.known_machines(self.tuning_dir) if mc.id != here.id]
 
     def machine_settings(self, machine_id=None):
-        """models.toml [machines.<id>] (optional): memory_headroom_gb, name."""
+        """models.toml [machines.<id>] (optional): memory_headroom_gb, gpu_memory_gb, name."""
         return self.cfg.get("machines", {}).get(machine_id or self.machine().id, {})
 
+    def headroom_gb(self, mc):
+        """GPU memory kept free: memory_headroom_gb, else 4 GB on a Mac (macOS and other apps share
+        it) and 1.5 GB in a discrete GPU's VRAM."""
+        return float(self.machine_settings(mc.id).get("memory_headroom_gb", 1.5 if mc.discrete else 4.0))
+
     def parallel_models(self):
         """How many models a run serves at a time on this machine: models.toml [machines.<id>]
         parallel_models (default 1)."""
@@ -1224,7 +1241,7 @@ class Engine:
     def memory_available_gb(self):
         """GPU memory models may take here (what the fit check sizes contexts against)."""
         mc = self.machine()
-        return mc.gpu_limit_gb - self.machine_settings(mc.id).get("memory_headroom_gb", 4.0)
+        return mc.gpu_limit_gb - self.headroom_gb(mc)
 
     def runs_alone(self, m):
         """A server whose before_start step may stop other servers never runs next to another model."""
@@ -1269,8 +1286,7 @@ class Engine:
                 if fit_check(server):
                     mmproj = expand(m.get("mmproj", ""))
                     extra = os.path.getsize(mmproj) if mmproj and os.path.isfile(mmproj) else 0
-                    headroom = self.machine_settings(mc.id).get("memory_headroom_gb", 4.0)
-                    f = machines.fit(path, mc, kv_type=kv or "f16", headroom_gb=headroom, extra_bytes=extra,
+                    f = machines.fit(path, mc, kv_type=kv or "f16", headroom_gb=self.headroom_gb(mc), extra_bytes=extra,
                                      want_ctx=want)
                     ctx, fits, note, need = f.max_ctx, f.fits, f.note, f.need_gb
             except (OSError, ValueError) as e:
@@ -1484,11 +1500,22 @@ class Engine:
                         time.sleep(0.2)
         threading.Thread(target=run, daemon=True).start()
 
-    def _watch_memory(self, proc, label, stop, peak):
+    VRAM_EVERY_S = 6   # how often to read a discrete GPU's VRAM while a server runs
+
+    def _watch_memory(self, proc, label, stop, peak, info=None):
         """Peak resident memory of the server's process group (MB). On Apple Silicon this includes
-        the model weights mapped for the GPU."""
+        the model weights mapped for the GPU. With discrete GPUs, also the peak VRAM it holds
+        (info["vram_peak_gb"]): its own processes' on Nvidia, the machine's on AMD."""
+        discrete = info is not None and self.machine().discrete
+        last_vram = [0.0]
+
         def run():
             while not stop.is_set():
+                if discrete and time.time() - last_vram[0] >= self.VRAM_EVERY_S:
+                    last_vram[0] = time.time()
+                    used = machines.vram_used_gb(machines.tree_pids(proc.pid))
+                    if used:
+                        info["vram_peak_gb"] = round(max(info.get("vram_peak_gb") or 0, used), 2)
                 try:
                     out = subprocess.run(["ps", "-axo", "pgid=,rss="], capture_output=True, text=True).stdout
                     rss = sum(int(r) for g, r in (l.split() for l in out.splitlines() if l.strip())
@@ -1789,7 +1816,7 @@ class Engine:
                                           f"(see logs/server/{log_name or m['label']}.log)")
                 t0 = time.time()
                 srv = self._start(cmd, log, cwd=cwd)
-                self._watch_memory(srv, m["label"], stop, peak)
+                self._watch_memory(srv, m["label"], stop, peak, info)
             else:
                 key_env = server.get("api_key_env")
                 if key_env and not os.environ.get(key_env):
@@ -1881,6 +1908,7 @@ class Engine:
             def run_info():
                 return {"load_s": info["load_s"], "peak_rss_mb": info["peak_mb"](),
                         "server_facts": info["facts"] or None, "gpu_peak_gb": info.get("gpu_peak_gb"),
+                        "vram_peak_gb": info.get("vram_peak_gb"),
                         "kernel_share_max": info.get("kernel_share_max"),
                         **({"loaded_alongside": loaded_alongside} if loaded_alongside else {}), **serving_info}
             failed_in_row = 0
@@ -1965,7 +1993,8 @@ class Engine:
         return tests
 
     # ---- sittings: a pack run over several sessions (stopped and resumed) keeps its total time
-    SITTING_KEYS = ("load_s", "peak_rss_mb", "gpu_peak_gb", "kernel_share_max", "machine", "loaded_alongside")
+    SITTING_KEYS = ("load_s", "peak_rss_mb", "gpu_peak_gb", "vram_peak_gb", "kernel_share_max", "machine",
+                    "loaded_alongside")
 
     def _sitting(self, job, info):
         now = time.time()
@@ -2037,7 +2066,7 @@ class Engine:
         out = {"started": min(s["started"] for s in sittings),
                "wall_s": round(sum(s.get("wall_s") or 0 for s in sittings), 1),
                "sittings": len(sittings), "sittings_log": sittings}
-        for k in ("peak_rss_mb", "gpu_peak_gb", "kernel_share_max"):
+        for k in ("peak_rss_mb", "gpu_peak_gb", "vram_peak_gb", "kernel_share_max"):
             if peak(k) is not None:
                 out[k] = peak(k)
         along = sorted({l for s in sittings for l in s.get("loaded_alongside") or []})

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/machines.py

@@ -1,6 +1,7 @@
 """Which machine this is, what a model needs, and whether it fits.
 
-    detect()              -> Machine (chip, cores, RAM, GPU memory limit, id like "m3max-64gb")
+    detect()              -> Machine (chip, cores, RAM, GPU memory limit, id like "m3max-64gb";
+                             with Nvidia/AMD cards on Linux, their VRAM and an id like "rtx4090-128gb")
     read_gguf(path)       -> GGUF header facts (architecture, layers, KV heads, context length, …)
     fit(model_path, …)    -> Fit (largest context that fits this machine, estimated memory)
 
@@ -9,6 +10,7 @@ the KV cache on its own (that would change answers); a quantized-KV variant is a
 """
 import dataclasses
 import functools
+import json
 import os
 import platform
 import re
@@ -28,15 +30,19 @@ class Machine:
     id: str
     chip: str
     ram_gb: float
-    gpu_limit_gb: float      # what Metal may wire for the GPU
+    gpu_limit_gb: float      # what Metal may wire for the GPU; with discrete GPUs, their VRAM (all cards)
     p_cores: int
     e_cores: int
     gpu_cores: int | None
     os: str
+    gpu_name: str | None = None   # discrete GPUs, e.g. "2× NVIDIA GeForce RTX 4090"
+    discrete: bool = False        # the GPU has its own memory (VRAM), separate from RAM
 
     @property
     def summary(self):
         cores = f"{self.p_cores}P+{self.e_cores}E" if self.e_cores else f"{self.p_cores} cores"
+        if self.discrete:
+            return f"{self.chip} · {self.ram_gb:.0f} GB RAM · {self.gpu_name} · {self.gpu_limit_gb:.0f} GB VRAM · {cores}"
         gpu = f", {self.gpu_cores}-core GPU" if self.gpu_cores else ""
         return f"{self.chip} · {self.ram_gb:.0f} GB (GPU up to {self.gpu_limit_gb:.0f} GB) · {cores}{gpu}"
 
@@ -53,6 +59,129 @@ def machine_id(chip, ram_gb):
     return re.sub(r"[^a-z0-9]+", "", name.lower()) + f"-{round(ram_gb)}gb"
 
 
+def gpu_id(gpus, ram_gb):
+    """A machine id from its discrete GPUs: "rtx4090-128gb", "2xrtx4090-128gb", "rx7900xtx+rtx3090-64gb"."""
+    names = []
+    for name in sorted({n for n, _ in gpus}):
+        short = re.sub(r"\b(nvidia|geforce|amd|radeon|ati|graphics|tm)\b", "", name.lower())
+        short = re.sub(r"[^a-z0-9]+", "", short) or "gpu"
+        count = sum(n == name for n, _ in gpus)
+        names.append(f"{count}x{short}" if count > 1 else short)
+    return "+".join(names) + f"-{round(ram_gb)}gb"
+
+
+# ---------------------------------------------------------------- discrete GPUs (Linux: Nvidia, AMD)
+# Read from the vendors' own tools. Anything unexpected counts as "not detected", never an error.
+def _run(cmd, timeout=10):
+    try:
+        p = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout)
+    except (OSError, subprocess.TimeoutExpired):
+        return None
+    return p.stdout if p.returncode == 0 else None
+
+
+def _nvidia_gpus():
+    out = _run(["nvidia-smi", "--query-gpu=name,memory.total", "--format=csv,noheader,nounits"])
+    gpus = []
+    for line in (out or "").splitlines():
+        name, _, mib = line.rpartition(",")
+        try:
+            gpus.append((name.strip(), float(mib) / 1024))
+        except ValueError:
+            pass
+    return [g for g in gpus if g[0] and g[1] > 0]
+
+
+def _walk(obj):
+    """Every dict inside a JSON value."""
+    if isinstance(obj, dict):
+        yield obj
+        for v in obj.values():
+            yield from _walk(v)
+    elif isinstance(obj, list):
+        for v in obj:
+            yield from _walk(v)
+
+
+def _bytes(value, unit=None):
+    """VRAM sizes as amd-smi/rocm-smi print them: 25753026560, "24560 MB", {"value": 24560, "unit": "MB"}."""
+    if isinstance(value, dict):
+        return _bytes(value.get("value"), value.get("unit"))
+    m = re.match(r"\s*([\d.]+)\s*([KMGT]i?B|B)?\s*$", str(value), re.I)
+    if not m:
+        return None
+    scale = {"B": 1, "KB": 2**10, "MB": 2**20, "GB": 2**30, "TB": 2**40}
+    u = (m.group(2) or unit or "B").upper().replace("I", "")
+    return float(m.group(1)) * scale.get(u, 1)
+
+
+def _amd_gpus():
+    out = _run(["amd-smi", "static", "--asic", "--vram", "--json"])
+    gpus = []
+    try:
+        for d in _walk(json.loads(out)) if out else []:
+            if "vram" in d and isinstance(d["vram"], dict):
+                size = _bytes(d["vram"].get("size"))
+                name = (d.get("asic") or {}).get("market_name") or "AMD GPU"
+                if size:
+                    gpus.append((str(name), size / GB))
+    except ValueError:
+        gpus = []
+    if gpus:
+        return gpus
+    out = _run(["rocm-smi", "--showmeminfo", "vram", "--showproductname", "--json"])
+    try:
+        cards = json.loads(out) if out else {}
+    except ValueError:
+        return []
+    for key, d in sorted(cards.items()) if isinstance(cards, dict) else []:
+        if not key.startswith("card") or not isinstance(d, dict):
+            continue
+        size = _bytes(d.get("VRAM Total Memory (B)"))
+        name = d.get("Card Series") or d.get("Card series") or d.get("Card SKU") or d.get("Card model") or "AMD GPU"
+        if size:
+            gpus.append((str(name), size / GB))
+    return gpus
+
+
+def discrete_gpus():
+    """[(name, VRAM GB)] for the Nvidia or AMD cards this machine has, or [] (none, or tools missing)."""
+    return _nvidia_gpus() or _amd_gpus()
+
+
+def vram_used_gb(pids=None):
+    """VRAM in use now (GB): by these processes on Nvidia (all of it when pids is None), or on the
+    whole machine on AMD, whose tools don't report it reliably per process. None if unknown."""
+    if pids is not None:
+        out = _run(["nvidia-smi", "--query-compute-apps=pid,used_memory", "--format=csv,noheader,nounits"])
+        if out is not None:
+            used = 0.0
+            for line in out.splitlines():
+                pid, _, mib = line.partition(",")
+                if pid.strip() in {str(p) for p in pids}:
+                    try:
+                        used += float(mib) / 1024
+                    except ValueError:
+                        pass
+            return used
+    out = _run(["nvidia-smi", "--query-gpu=memory.used", "--format=csv,noheader,nounits"])
+    if out is not None:
+        try:
+            return sum(float(x) for x in out.split()) / 1024
+        except ValueError:
+            return None
+    out = _run(["rocm-smi", "--showmeminfo", "vram", "--json"])
+    try:
+        cards = json.loads(out) if out else None
+    except ValueError:
+        return None
+    if not isinstance(cards, dict):
+        return None
+    used = [_bytes(d.get("VRAM Total Used Memory (B)")) for k, d in cards.items()
+            if k.startswith("card") and isinstance(d, dict)]
+    return sum(u for u in used if u) / GB if any(used) else None
+
+
 @functools.lru_cache(maxsize=1)
 def detect():
     """This machine. Override the id with EVALS_MACHINE if two machines would collide."""
@@ -73,11 +202,19 @@ def detect():
         except (OSError, subprocess.TimeoutExpired):
             pass
         os_name = "macOS " + platform.mac_ver()[0]
-    else:  # Linux and others: CPU-side facts only
+    else:  # Linux and others: the CPU side, plus Nvidia/AMD cards if their tools are installed
         chip = platform.processor() or platform.machine()
         ram = os.sysconf("SC_PAGE_SIZE") * os.sysconf("SC_PHYS_PAGES") / GB
         p, e, gpu_cores, limit = os.cpu_count() or 1, 0, None, ram * 0.9
         os_name = platform.system()
+        gpus = discrete_gpus()
+        if gpus:
+            names = sorted({n for n, _ in gpus})
+            gpu_name = " + ".join(f"{sum(n == x for n, _ in gpus)}× {x}" if sum(n == x for n, _ in gpus) > 1 else x
+                                  for x in names)
+            mid = os.environ.get("EVALS_MACHINE") or gpu_id(gpus, ram)
+            return Machine(mid, chip, round(ram, 1), round(sum(g for _, g in gpus), 1), p, e, None, os_name,
+                           gpu_name=gpu_name, discrete=True)
     mid = os.environ.get("EVALS_MACHINE") or machine_id(chip, ram)
     return Machine(mid, chip, round(ram, 1), round(limit, 1), p, e, gpu_cores, os_name)
 
@@ -195,14 +332,15 @@ def fit(model_path, machine=None, kv_type="f16", headroom_gb=4.0, extra_bytes=0,
     model_max = info["context_length"] or 131072
     max_ctx = int(min(room / per_tok, model_max, want_ctx or model_max)) if room > 0 else 0
     max_ctx = max_ctx // 1024 * 1024
+    of = " of VRAM" if machine.discrete else ""
     if max_ctx < MIN_CTX:
         need = fixed + MIN_CTX * per_tok / GB
         return Fit(False, max_ctx, round(need, 1), round(available, 1),
                    f"doesn't fit on {machine.id}: needs ~{need:.0f} GB at {MIN_CTX // 1024}k context, "
-                   f"{available:.0f} GB available")
+                   f"{available:.0f} GB{of} available")
     need = fixed + max_ctx * per_tok / GB
     return Fit(True, max_ctx, round(need, 1), round(available, 1),
-               f"fits on {machine.id} up to {max_ctx // 1024}k context (~{need:.0f} of {available:.0f} GB)")
+               f"fits on {machine.id} up to {max_ctx // 1024}k context (~{need:.0f} of {available:.0f} GB{of})")
 
 
 # ---------------------------------------------------------------- memory pressure
@@ -287,6 +425,26 @@ def tree_cpu_seconds(pid):
     return (sum(rows[p][1] for p in tree), sum(rows[p][2] for p in tree))
 
 
+def tree_pids(pid):
+    """A process and all its descendants (pids), or [pid] if ps can't be read."""
+    try:
+        out = subprocess.run(["ps", "-axo", "pid=,ppid="], capture_output=True, text=True, timeout=5).stdout
+    except (OSError, subprocess.TimeoutExpired):
+        return [pid]
+    children = {}
+    for line in out.splitlines():
+        parts = line.split()
+        if len(parts) == 2:
+            children.setdefault(parts[1], []).append(parts[0])
+    tree, frontier = {str(pid)}, [str(pid)]
+    while frontier:
+        for child in children.get(frontier.pop(), []):
+            if child not in tree:
+                tree.add(child)
+                frontier.append(child)
+    return sorted(int(p) for p in tree)
+
+
 def group_cpu_seconds(pgid):
     """(user, kernel) CPU seconds used so far by a process group, or None."""
     try:

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/src/tuieval/tune.py

@@ -60,6 +60,7 @@ SWAP_NOTABLE = 256 * 2**20  # swap worth a warning (bytes); rejecting for swap n
 ANSWER_TOKENS = 1024       # answer length the projected objective scores ([tune] answer_tokens)
 PRESSURE_CRITICAL = 4      # macOS memory pressure level that rejects a candidate
 PRESSURE_EVERY_S = 2
+VRAM_MARGIN_GB = 0.25     # a discrete GPU this close to full could run out on a longer prompt
 GPU_MARGIN_GB = 0.75      # a candidate whose GPU allocation comes this close to the residency limit
                           # is rejected even before it stalls: servers grow as requests arrive
 REQUEST_LIMIT_S = 600     # a tuning request taking longer fails the candidate ([tune] request_timeout_s)
@@ -556,6 +557,14 @@ def tune(eng, label, emit=lambda *a, **k: None, max_starts=16, min_gain=0.03, us
                     r.facts["gpu_settled_gb"] = round(settled, 1)
                 if info.get("kernel_share_max") is not None:
                     r.facts["kernel_share"] = info["kernel_share_max"]
+                mc = eng.machine()
+                if mc.discrete:   # VRAM is the limit there: its own memory, nothing else can use it
+                    vram = machines.vram_used_gb()
+                    if vram is not None:
+                        r.facts["vram_used_gb"] = round(vram, 1)
+                        if vram > mc.gpu_limit_gb - VRAM_MARGIN_GB and not r.error:
+                            r.error = (f"VRAM nearly full: {vram:.1f} of {mc.gpu_limit_gb:.0f} GB in use "
+                                       "(a longer prompt could run out)")
                 limit = eng.gpu_residency_gb()
                 if settled and settled > limit - GPU_MARGIN_GB and not r.error:
                     r.error = (f"GPU allocation settled at {settled:.1f} GB, within {GPU_MARGIN_GB} GB of the "

{tuieval-0.2.0.dev62 → tuieval-0.2.0.dev65}/tests/test_tuieval.py

@@ -902,6 +902,153 @@ class Unavailable(unittest.TestCase):
             self.assertIn("works", tuieval(ws, "pta").stdout)
 
 
+class DiscreteGPU(unittest.TestCase):
+    """Nvidia and AMD cards on Linux, read from their vendors' tools (faked here, so no GPU is needed)."""
+
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.bin = os.path.join(self.tmp.name, "bin")
+        os.makedirs(self.bin)
+        self.path = os.environ["PATH"]
+        os.environ["PATH"] = self.bin + os.pathsep + "/usr/bin:/bin"   # only the fakes, never a real GPU tool
+
+    def tearDown(self):
+        os.environ["PATH"] = self.path
+        self.tmp.cleanup()
+
+    def tool(self, name, script):
+        path = os.path.join(self.bin, name)
+        write(path, "#!/bin/sh\n" + script)
+        os.chmod(path, 0o755)
+
+    def test_nvidia(self):
+        from tuieval import machines
+        self.tool("nvidia-smi", textwrap.dedent("""
+            case "$*" in
+              *query-gpu=name,memory.total*) printf 'NVIDIA GeForce RTX 4090, 24564\nNVIDIA GeForce RTX 4090, 24564\n' ;;
+              *query-compute-apps*) printf '4242, 18000\n999, 3000\n' ;;
+              *query-gpu=memory.used*) printf '18500\n200\n' ;;
+            esac
+        """))
+        gpus = machines.discrete_gpus()
+        self.assertEqual([n for n, _ in gpus], ["NVIDIA GeForce RTX 4090"] * 2)
+        self.assertAlmostEqual(sum(g for _, g in gpus), 47.98, places=1)
+        self.assertEqual(machines.gpu_id(gpus, 127.6), "2xrtx4090-128gb")
+        self.assertAlmostEqual(machines.vram_used_gb([4242, 4243]), 18000 / 1024)   # the server's own
+        self.assertAlmostEqual(machines.vram_used_gb(), 18700 / 1024)              # the whole machine
+
+    def test_amd(self):
+        from tuieval import machines
+        self.tool("amd-smi", "exit 1\n")                                         # older ROCm: rocm-smi
+        self.tool("rocm-smi", textwrap.dedent("""
+            case "$*" in
+              *showproductname*) echo '{"card0": {"VRAM Total Memory (B)": "25753026560", "Card Series": "Radeon RX 7900 XTX"}, "system": {}}' ;;
+              *) echo '{"card0": {"VRAM Total Used Memory (B)": "4294967296"}}' ;;
+            esac
+        """))
+        self.assertEqual(machines.discrete_gpus(), [("Radeon RX 7900 XTX", 25753026560 / 2**30)])
+        self.assertEqual(machines.gpu_id(machines.discrete_gpus(), 64), "rx7900xtx-64gb")
+        self.assertAlmostEqual(machines.vram_used_gb([1]), 4.0)                    # machine-wide on AMD
+        self.tool("amd-smi", "echo '[{\"gpu\": 0, \"asic\": {\"market_name\": \"Radeon PRO W7900\"}, "
+                             "\"vram\": {\"size\": {\"value\": 49152, \"unit\": \"MB\"}}}]'\n")
+        self.assertEqual(machines.discrete_gpus(), [("Radeon PRO W7900", 48.0)])
+
+    def test_missing_or_broken_tools_mean_no_gpu(self):
+        from tuieval import machines
+        self.assertEqual(machines.discrete_gpus(), [])
+        self.assertIsNone(machines.vram_used_gb([1]))
+        self.tool("nvidia-smi", "echo 'NVIDIA-SMI has failed because it could not communicate with the driver'\nexit 9\n")
+        self.tool("rocm-smi", "echo 'not json'\n")
+        self.assertEqual(machines.discrete_gpus(), [])
+
+    def test_sizing_ids_and_overrides(self):
+        from tuieval import engine, machines
+        gpu = machines.Machine("rtx4090-128gb", "x86_64", 128.0, 24.0, 16, 0, None, "Linux",
+                               gpu_name="NVIDIA GeForce RTX 4090", discrete=True)
+        self.assertIn("NVIDIA GeForce RTX 4090 · 24 GB VRAM", gpu.summary)
+        ws = os.path.join(self.tmp.name, "ws")
+        tuieval(self.tmp.name, "init", ws)
+        real = machines.detect
+        machines.detect = lambda: gpu
+        try:
+            e = engine.Engine(root=ws)
+            self.assertEqual((e.machine().id, e.headroom_gb(e.machine())), ("rtx4090-128gb", 1.5))
+            self.assertEqual(e.memory_available_gb(), 22.5)                       # VRAM, not 128 GB of RAM
+            os.makedirs(os.path.join(ws, "tuning", "x8664-128gb"))                # recorded before detection
+            self.assertEqual(e.machine().id, "x8664-128gb")                        # keeps its results' id
+            with open(os.path.join(ws, "models.toml"), "a") as f:
+                f.write("\n[machines.x8664-128gb]\ngpu_memory_gb = 20\nmemory_headroom_gb = 2\n")
+            e = engine.Engine(root=ws)
+            self.assertEqual((e.machine().gpu_limit_gb, e.memory_available_gb()), (20.0, 18.0))
+        finally:
+            machines.detect = real
+    def test_doctor_and_vram_peak(self):
+        import contextlib, io, threading
+        from tuieval import doctor, engine, machines
+        self.tool("nvidia-smi", "printf 'NVIDIA GeForce RTX 4090, 24564\\n'\n")
+        gpu = machines.Machine("x8664-128gb", "x86_64", 128.0, 24.0, 16, 0, None, "Linux",
+                               gpu_name="NVIDIA GeForce RTX 4090", discrete=True)
+        ws = os.path.join(self.tmp.name, "ws")
+        tuieval(self.tmp.name, "init", ws)
+        real, real_vram = machines.detect, machines.vram_used_gb
+        machines.detect = lambda: gpu
+        env = dict(os.environ)
+        os.environ.update(TUIEVAL_HOME=ws, TUIEVAL_DETECT_PORTS="", NO_COLOR="1")
+        try:
+            out = io.StringIO()
+            with contextlib.redirect_stdout(out):
+                doctor.check()
+            text = out.getvalue()
+            self.assertIn("NVIDIA GeForce RTX 4090 · 24 GB VRAM", text)
+            self.assertIn("sized to 24 GB of VRAM, keeping 1.5 GB free", text)
+            self.assertIn("EVALS_MACHINE=rtx4090-128gb", text)                    # an id kept from before
+            machines.vram_used_gb = lambda pids=None: 12.5
+            e, info, stop = engine.Engine(root=ws), {}, threading.Event()
+            proc = subprocess.Popen(["sleep", "5"])
+            try:
+                e._watch_memory(proc, "m", stop, [0.0], info)
+                time.sleep(1)
+            finally:
+                stop.set()
+                proc.kill()
+                proc.wait()
+            self.assertEqual(info.get("vram_peak_gb"), 12.5)
+        finally:
+            machines.detect, machines.vram_used_gb = real, real_vram
+            os.environ.clear()
+            os.environ.update(env)
+
+    def test_fit_messages_name_vram(self):
+        from tuieval import machines
+        path = os.path.join(self.tmp.name, "m.gguf")
+        fake_gguf(path, size=1000)
+        gpu = lambda gb: machines.Machine("rtx4090-128gb", "x86_64", 128.0, gb, 16, 0, None, "Linux",
+                                          gpu_name="NVIDIA GeForce RTX 4090", discrete=True)
+        ok, no = machines.fit(path, gpu(24.0), headroom_gb=1.5), machines.fit(path, gpu(1.2), headroom_gb=0.1)
+        self.assertTrue(ok.fits and ok.note.endswith("GB of VRAM)"), ok.note)
+        self.assertTrue(not no.fits and no.note.endswith("GB of VRAM available"), no.note)
+        mac = machines.Machine("m3max-64gb", "Apple M3 Max", 64.0, 48.0, 12, 4, 40, "macOS")
+        self.assertNotIn("VRAM", machines.fit(path, mac).note)
+
+    def test_vram_peak_reaches_the_pta_index(self):
+        from tuieval import engine, pta
+        sittings = [{"started": 1, "wall_s": 10, "vram_peak_gb": 18.2, "peak_rss_mb": 2048},
+                    {"started": 2, "wall_s": 10, "vram_peak_gb": 21.5, "peak_rss_mb": 1900}]
+        run = engine.Engine.sittings_summary(sittings)
+        self.assertEqual(run["vram_peak_gb"], 21.5)                   # the peak over every sitting is kept
+        infos = [{"label": "gpu", "vram_peak_gb": run["vram_peak_gb"], "peak_rss_mb": run["peak_rss_mb"]},
+                 {"label": "mac", "peak_rss_mb": 25600}]
+        self.assertEqual(pta.memory_gb(infos), {"gpu": 21.5, "mac": 25.0})   # VRAM where it was measured
+
+    def test_mac_is_unchanged(self):
+        from tuieval import engine, machines
+        e = engine.Engine.__new__(engine.Engine)
+        e.cfg = {"machines": {}}
+        mac = machines.Machine("m3max-64gb", "Apple M3 Max", 64.0, 48.0, 12, 4, 40, "macOS")
+        self.assertFalse(mac.discrete)                                             # Macs are unchanged
+        self.assertEqual(e.headroom_gb(mac), 4.0)
+
+
 class FirstRun(unittest.TestCase):
     """A new user's first run fails fast with the fix, never hangs or shows a bare Python error."""
 
@@ -1201,7 +1348,7 @@ class WarmTune(unittest.TestCase):
         fake_gguf(f"{d}/new.gguf", size=1000)
         fake_gguf(f"{d}/sib.gguf", size=1100)
         fake_gguf(f"{d}/other.gguf", embedding=4096, size=1000)
-        machine = types.SimpleNamespace(id="mac", summary="a Mac")
+        machine = types.SimpleNamespace(id="mac", summary="a Mac", discrete=False)
         eng = types.SimpleNamespace(
             cfg={"servers": {"llama": {"cmd": ["llama"], "tune": self.KNOBS}},
                  "models": [{"label": n, "server": "llama", "model": f"{d}/{n}.gguf"}
@@ -1212,6 +1359,7 @@ class WarmTune(unittest.TestCase):
         eng.serving = lambda m: types.SimpleNamespace(fits=True, ctx=8192, machine=machine, mtp_layers=1,
                                                       identity={"model_bytes": 1000})
         eng.knob_values = lambda mc: {}
+        eng.machine = lambda: machine
         eng.server_version = lambda m: "v1"
         eng.gpu_residency_gb = lambda: 24