The portal in English and Turkish: a catalogue per language, a switcher at the foot of every page, landing.language, Turkish numbers and dates
Every template and handler message in internal/web goes through a JSON
catalogue under internal/web/locales/<code>/, with English as the fallback.
The language is the person's choice (a cookie set by /_lang/{code}), else
Accept-Language, else landing.language. Tests refuse a key some language
lacks, a translation with other arguments than English, and template text
left in English.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
1 parent
30f7f3474a
commit
11231d8c2b
113 files changed
+8091
-1981
No files matched your search
+57
-56
@@ -86,7 +86,7 @@ type serveForm struct {
|
||||
}
|
||||
|
||||
// planServe works out what serving this version would mean.
|
||||
func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
func planServe(l *Lang, st mirror.Status, files []mirror.Status, base string,
|
||||
fleet Source, cat CatalogAdmin, check func(source string) gguf.Verdict,
|
||||
) serveForm {
|
||||
var header gguf.Header
|
||||
@@ -123,14 +123,14 @@ func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
|
||||
switch f.Engine {
|
||||
case "":
|
||||
f.Refusal = "nothing here reads " + extOf(st.File) + " files"
|
||||
f.Refusal = l.T("serve.refuse.no_engine", extOf(st.File))
|
||||
return f
|
||||
case "llamacpp":
|
||||
f.Profile = "cpu"
|
||||
if check != nil {
|
||||
v := check(f.Source)
|
||||
if !v.Servable {
|
||||
f.Refusal = "llama.cpp on the nodes cannot run this: " + v.Reason
|
||||
f.Refusal = l.T("serve.refuse.llamacpp", v.Reason)
|
||||
return f
|
||||
}
|
||||
header = v.Header
|
||||
@@ -139,7 +139,7 @@ func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
f.Profile = "gpu"
|
||||
}
|
||||
f.GPUs, f.Parallel = 1, defaultParallel
|
||||
f.Placements, f.FreeGPUs = placementsFor(fleet, f.Engine)
|
||||
f.Placements, f.FreeGPUs = placementsFor(l, fleet, f.Engine)
|
||||
for _, pc := range f.Placements {
|
||||
if pc.Selected {
|
||||
f.Profile = pc.profile()
|
||||
@@ -152,15 +152,13 @@ func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
// A warning, not a refusal: an engine can be installed this afternoon, and
|
||||
// a node that reports nothing is an older agent rather than an empty one.
|
||||
if !runsAnywhere(fleet, f.Engine) {
|
||||
f.Warning = "no node reports having " + f.Engine + " installed, and " +
|
||||
extOf(st.File) + " weights need it"
|
||||
f.Warning = l.T("serve.warn.not_installed", f.Engine, extOf(st.File))
|
||||
if f.Engine != "llamacpp" && gpus(fleet) == 0 {
|
||||
f.Warning += ". This fleet has no GPU either; a GGUF build of the " +
|
||||
"same model runs on the CPU nodes"
|
||||
f.Warning += ". " + l.T("serve.warn.no_gpu")
|
||||
}
|
||||
}
|
||||
f.CtxPool = defaultCtxPool
|
||||
f.MinMemoryGB, f.MemoryBasis = memoryFor(st.BytesTotal, st.BytesDone), "weights plus half again plus two"
|
||||
f.MinMemoryGB, f.MemoryBasis = memoryFor(st.BytesTotal, st.BytesDone), l.T("serve.basis.estimate")
|
||||
weights := st.BytesTotal
|
||||
if weights <= 0 {
|
||||
weights = st.BytesDone
|
||||
@@ -175,10 +173,10 @@ func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
// across the models on the live mirror at the same 16k — a spread the
|
||||
// flat "half again" could not see.
|
||||
f.MinMemoryGB = int((b + 1<<30 - 1) >> 30)
|
||||
f.MemoryBasis = fmt.Sprintf("weights %.1f GB + KV cache %.1f GB for %d tokens (%d layers × %d KV heads) + %.1f GB for the engine",
|
||||
float64(weights)/(1<<30), float64(int64(defaultCtxPool)*header.KVBytesPerToken())/(1<<30),
|
||||
defaultCtxPool, header.BlockCount, kvHeads,
|
||||
float64(int64(1<<30)+weights/20)/(1<<30))
|
||||
f.MemoryBasis = l.T("serve.basis.shape",
|
||||
l.Num(float64(weights)/(1<<30), 1), l.Num(float64(int64(defaultCtxPool)*header.KVBytesPerToken())/(1<<30), 1),
|
||||
l.Int(defaultCtxPool), header.BlockCount, kvHeads,
|
||||
l.Num(float64(int64(1<<30)+weights/20)/(1<<30), 1))
|
||||
}
|
||||
if fleet != nil {
|
||||
var cpus []int
|
||||
@@ -190,14 +188,15 @@ func planServe(st mirror.Status, files []mirror.Status, base string,
|
||||
f.CPUCores = catalog.SuggestCores(cpus)
|
||||
}
|
||||
if biggest := biggestNode(fleet); biggest > 0 && f.MinMemoryGB > biggest {
|
||||
f.Warning = strings.TrimSuffix(f.Warning, ".") +
|
||||
fmt.Sprintf("%sthe largest node has %d GB and this needs about %d",
|
||||
sep(f.Warning), biggest, f.MinMemoryGB)
|
||||
big := l.T("serve.warn.too_big", biggest, f.MinMemoryGB)
|
||||
if f.Warning != "" {
|
||||
big = l.T("serve.warn.and", strings.TrimSuffix(f.Warning, "."), big)
|
||||
}
|
||||
f.Warning = big
|
||||
}
|
||||
if cat != nil && cat.Declared(st.Model) {
|
||||
f.Already = true
|
||||
f.Refusal = "a model called " + st.Model +
|
||||
" is declared in configuration; edit it there rather than adding a second"
|
||||
f.Refusal = l.T("serve.refuse.declared", st.Model)
|
||||
}
|
||||
return f
|
||||
}
|
||||
@@ -255,8 +254,9 @@ func gpus(fleet Source) int {
|
||||
return n
|
||||
}
|
||||
|
||||
// buildModel turns the submitted form into a catalogue entry.
|
||||
func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
// buildModel turns the submitted form into a catalogue entry. Its refusals
|
||||
// are in l, the language of whoever filled the form in.
|
||||
func buildModel(l *Lang, get func(string) string) (catalog.Model, error) {
|
||||
m := catalog.Model{
|
||||
Name: strings.TrimSpace(get("name")),
|
||||
Description: strings.TrimSpace(get("description")),
|
||||
@@ -264,7 +264,7 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
RoutingProfile: "chat",
|
||||
}
|
||||
if m.Name == "" || m.Source == "" {
|
||||
return m, errors.New("a model needs a name and a source")
|
||||
return m, errors.New(l.T("serve.err.name_source"))
|
||||
}
|
||||
for _, a := range strings.Split(get("aliases"), ",") {
|
||||
if a = strings.TrimSpace(a); a != "" {
|
||||
@@ -272,18 +272,18 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
}
|
||||
}
|
||||
|
||||
ctxLen, err := positive(get("context_length"), "the context length")
|
||||
ctxLen, err := positive(l, get("context_length"), "serve.what.context_length")
|
||||
if err != nil {
|
||||
return m, err
|
||||
}
|
||||
m.ContextLength = ctxLen
|
||||
if v := strings.TrimSpace(get("max_output")); v != "" {
|
||||
if m.MaxOutput, err = positive(v, "the output limit"); err != nil {
|
||||
if m.MaxOutput, err = positive(l, v, "serve.what.max_output"); err != nil {
|
||||
return m, err
|
||||
}
|
||||
}
|
||||
|
||||
min, err := positive(get("replicas"), "the replica count")
|
||||
min, err := positive(l, get("replicas"), "serve.what.replicas")
|
||||
if err != nil {
|
||||
return m, err
|
||||
}
|
||||
@@ -291,12 +291,12 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
|
||||
engine := strings.TrimSpace(get("engine"))
|
||||
if engine == "" {
|
||||
return m, errors.New("which engine should run it?")
|
||||
return m, errors.New(l.T("serve.err.engine"))
|
||||
}
|
||||
p := catalog.Profile{Engine: engine, Weight: 1, Priority: 100}
|
||||
gpus := 1
|
||||
if v := strings.TrimSpace(get("gpus")); v != "" {
|
||||
if gpus, err = positive(v, "the GPUs per replica"); err != nil {
|
||||
if gpus, err = positive(l, v, "serve.what.gpus"); err != nil {
|
||||
return m, err
|
||||
}
|
||||
gpus = max(gpus, 1)
|
||||
@@ -316,11 +316,11 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
place.GPUs = gpus
|
||||
case place.Value == "" || place.Value == "cpu":
|
||||
if engine != "llamacpp" {
|
||||
return m, fmt.Errorf("%s runs on GPUs; choose a GPU class to place it on", engine)
|
||||
return m, errors.New(l.T("serve.err.needs_gpu", engine))
|
||||
}
|
||||
place.Value = "cpu"
|
||||
default:
|
||||
return m, fmt.Errorf("%q is not a placement", place.Value)
|
||||
return m, errors.New(l.T("serve.err.placement", place.Value))
|
||||
}
|
||||
p.Name = place.profile()
|
||||
if v := strings.TrimSpace(get("profile")); v != "" {
|
||||
@@ -330,13 +330,13 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
// parallel slots, so the number a client may use is smaller and is what
|
||||
// context_length above says.
|
||||
if v := strings.TrimSpace(get("ctx_size")); v != "" {
|
||||
if _, err := positive(v, "the engine's context pool"); err != nil {
|
||||
if _, err := positive(l, v, "serve.what.ctx_pool"); err != nil {
|
||||
return m, err
|
||||
}
|
||||
p.Args = []string{"--ctx-size", v}
|
||||
}
|
||||
if v := strings.TrimSpace(get("parallel")); v != "" && engine == "llamacpp" {
|
||||
n, err := positive(v, "the parallel requests")
|
||||
n, err := positive(l, v, "serve.what.parallel")
|
||||
if err != nil {
|
||||
return m, err
|
||||
}
|
||||
@@ -345,12 +345,12 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
}
|
||||
}
|
||||
if v := strings.TrimSpace(get("min_memory_gb")); v != "" {
|
||||
if p.MinMemoryGB, err = positive(v, "the memory requirement"); err != nil {
|
||||
if p.MinMemoryGB, err = positive(l, v, "serve.what.memory"); err != nil {
|
||||
return m, err
|
||||
}
|
||||
}
|
||||
if v := strings.TrimSpace(get("cpu_cores")); v != "" {
|
||||
if p.CPUCores, err = positive(v, "the cores"); err != nil {
|
||||
if p.CPUCores, err = positive(l, v, "serve.what.cores"); err != nil {
|
||||
return m, err
|
||||
}
|
||||
}
|
||||
@@ -364,14 +364,22 @@ func buildModel(get func(string) string) (catalog.Model, error) {
|
||||
return m, nil
|
||||
}
|
||||
|
||||
func positive(v, what string) (int, error) {
|
||||
// positive reads a whole number from a form field; what is the key naming
|
||||
// the field, for the refusal.
|
||||
func positive(l *Lang, v, what string) (int, error) {
|
||||
n, err := strconv.Atoi(strings.TrimSpace(v))
|
||||
if err != nil || n < 0 {
|
||||
return 0, fmt.Errorf("%s must be a whole number", what)
|
||||
return 0, errors.New(l.T("serve.err.whole_number", l.T(what)))
|
||||
}
|
||||
return n, nil
|
||||
}
|
||||
|
||||
func init() {
|
||||
registerKeys("serve.what.context_length", "serve.what.max_output", "serve.what.replicas",
|
||||
"serve.what.gpus", "serve.what.ctx_pool", "serve.what.parallel", "serve.what.memory",
|
||||
"serve.what.cores")
|
||||
}
|
||||
|
||||
// memoryFor estimates the host RAM a model needs, from the size of its weights.
|
||||
//
|
||||
// Weights plus half again, plus two. The half covers the KV pool, which grows
|
||||
@@ -409,14 +417,6 @@ func biggestNode(fleet Source) int {
|
||||
return most
|
||||
}
|
||||
|
||||
// sep joins a second clause onto a warning that may be empty.
|
||||
func sep(existing string) string {
|
||||
if existing == "" {
|
||||
return ""
|
||||
}
|
||||
return ", and "
|
||||
}
|
||||
|
||||
// Run modes a page offers for a model, in the words a person would use.
|
||||
const (
|
||||
modeAlways = "always"
|
||||
@@ -497,7 +497,7 @@ func (c placeChoice) profile() string {
|
||||
// on a cluster of GPUs, ran at a tenth of their speed, and nothing on the way
|
||||
// said why — the form had no choice to offer and named the profile "cpu". The
|
||||
// class selected is the one with the most cards free.
|
||||
func placementsFor(fleet Source, engine string) ([]placeChoice, int) {
|
||||
func placementsFor(l *Lang, fleet Source, engine string) ([]placeChoice, int) {
|
||||
var out []placeChoice
|
||||
free, best := 0, -1
|
||||
if fleet != nil {
|
||||
@@ -506,7 +506,7 @@ func placementsFor(fleet Source, engine string) ([]placeChoice, int) {
|
||||
free += n
|
||||
out = append(out, placeChoice{
|
||||
Value: "gpu:" + c.Class,
|
||||
Label: fmt.Sprintf("GPU: %s, %d of %d free", c.Class, n, c.Total),
|
||||
Label: l.T("serve.place.gpu", c.Class, n, c.Total),
|
||||
})
|
||||
if best < 0 || n > fleet.Summary().GPUs[best].Free() {
|
||||
best = len(out) - 1
|
||||
@@ -517,9 +517,9 @@ func placementsFor(fleet Source, engine string) ([]placeChoice, int) {
|
||||
out[best].Selected = true
|
||||
}
|
||||
if engine == "llamacpp" {
|
||||
label := "CPU only: no GPU, many times slower"
|
||||
label := l.T("serve.place.cpu_slower")
|
||||
if len(out) == 0 {
|
||||
label = "CPU: this fleet has no GPU"
|
||||
label = l.T("serve.place.cpu_only")
|
||||
}
|
||||
out = append(out, placeChoice{Value: "cpu", Label: label, Selected: best < 0})
|
||||
}
|
||||
@@ -539,8 +539,9 @@ func freeGPUs(fleet Source) int {
|
||||
}
|
||||
|
||||
// describeServe says, in one line, what a model will run as: the summary
|
||||
// shown before it is added and in the message after.
|
||||
func describeServe(m catalog.Model) string {
|
||||
// shown before it is added and in the message after. The form's script says
|
||||
// the same with the same sentences, so the two cannot drift apart in wording.
|
||||
func describeServe(l *Lang, m catalog.Model) string {
|
||||
if len(m.Profiles) == 0 {
|
||||
return ""
|
||||
}
|
||||
@@ -549,27 +550,27 @@ func describeServe(m catalog.Model) string {
|
||||
if engine == "" {
|
||||
engine = p.Engine
|
||||
}
|
||||
where := "on the CPU"
|
||||
where := l.T("serve.on_cpu", engine)
|
||||
if p.GPUClass != "" {
|
||||
n := max(p.GPUsPerNode, p.TP, 1)
|
||||
where = fmt.Sprintf("on %s %s", plural(n, "GPU", "GPUs"), p.GPUClass)
|
||||
where = l.T("serve.on_gpu_one", engine, p.GPUClass)
|
||||
if n > 1 {
|
||||
where = fmt.Sprintf("on %d %s GPUs of one node", n, p.GPUClass)
|
||||
where = l.T("serve.on_gpus", engine, n, p.GPUClass)
|
||||
}
|
||||
}
|
||||
parts := []string{engine + " " + where}
|
||||
parts := []string{where}
|
||||
if p.MinMemoryGB > 0 {
|
||||
parts = append(parts, fmt.Sprintf("%d GB of host memory", p.MinMemoryGB))
|
||||
parts = append(parts, l.T("serve.host_memory", l.Int(int64(p.MinMemoryGB))))
|
||||
}
|
||||
if p.GPUClass == "" && p.CPUCores > 0 {
|
||||
parts = append(parts, plural(p.CPUCores, "core", "cores"))
|
||||
parts = append(parts, l.N("serve.cores", int64(p.CPUCores)))
|
||||
}
|
||||
pool, parallel := argInt(p.Args, "--ctx-size"), argInt(p.Args, "--parallel")
|
||||
switch {
|
||||
case pool > 0 && parallel > 0:
|
||||
parts = append(parts, fmt.Sprintf("context pool %d over %d parallel requests (%d each)", pool, parallel, pool/parallel))
|
||||
parts = append(parts, l.T("serve.pool_parallel", l.Int(int64(pool)), l.Int(int64(parallel)), l.Int(int64(pool/parallel))))
|
||||
case pool > 0:
|
||||
parts = append(parts, fmt.Sprintf("context pool %d", pool))
|
||||
parts = append(parts, l.T("serve.pool", l.Int(int64(pool))))
|
||||
}
|
||||
return strings.Join(parts, ", ")
|
||||
}
|
||||
|
||||
Reference in new issue
Block a user