Skip to content

Commit 9067bc8

Browse files
authored
Updates for the new release of kv-cache-manager (#194)
Signed-off-by: Ira <[email protected]>
1 parent 2bcfedb commit 9067bc8

File tree

5 files changed

+12
-34
lines changed

5 files changed

+12
-34
lines changed

go.mod

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8,7 +8,7 @@ require (
88
github.com/buaazp/fasthttprouter v0.1.1
99
github.com/go-logr/logr v1.4.2
1010
github.com/google/uuid v1.6.0
11-
github.com/llm-d/llm-d-kv-cache-manager v0.2.1
11+
github.com/llm-d/llm-d-kv-cache-manager v0.3.0-rc1
1212
github.com/onsi/ginkgo/v2 v2.23.4
1313
github.com/onsi/gomega v1.37.0
1414
github.com/openai/openai-go v0.1.0-beta.10

go.sum

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,5 @@
1+
github.com/alicebob/miniredis/v2 v2.35.0 h1:QwLphYqCEAo1eu1TqPRN2jgVMPBweeQcR21jeqDCONI=
2+
github.com/alicebob/miniredis/v2 v2.35.0/go.mod h1:TcL7YfarKPGDAthEtl5NBeHZfeUQj6OXMm/+iu5cLMM=
13
github.com/andybalholm/brotli v1.1.1 h1:PR2pgnyFznKEugtsUo0xLdDop5SKXd5Qf5ysW+7XdTA=
24
github.com/andybalholm/brotli v1.1.1/go.mod h1:05ib4cKhjx3OQYUY22hTVd34Bc8upXjOLL2rKwwZBoA=
35
github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
@@ -66,8 +68,8 @@ github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
6668
github.com/kr/text v0.2.0/go.mod h1:eLer722TekiGuMkidMxC/pM04lWEeraHUUmBw8l2grE=
6769
github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0SNc=
6870
github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
69-
github.com/llm-d/llm-d-kv-cache-manager v0.2.1 h1:PKIjJPUF9ILLFBNvZRa0QQ/liTQjBKwWChzcenEdM08=
70-
github.com/llm-d/llm-d-kv-cache-manager v0.2.1/go.mod h1:s1xaE4ImkihWaLg2IQh4VN6L1PgN5RD1u1VarPey6dw=
71+
github.com/llm-d/llm-d-kv-cache-manager v0.3.0-rc1 h1:SDLiNrcreDcA9m9wfXAumFARDHHXpjOjHTzshTiTGxk=
72+
github.com/llm-d/llm-d-kv-cache-manager v0.3.0-rc1/go.mod h1:tN80/D0Faf6pE2ocwFgTNoCxKPsqdsa2XnjQUqOaZ8Q=
7173
github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0=
7274
github.com/mailru/easyjson v0.7.7/go.mod h1:xzfreul335JAWq5oZzymOObrkdz5UnU4kGfJJLY9Nlc=
7375
github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
@@ -143,6 +145,8 @@ github.com/xyproto/randomstring v1.0.5 h1:YtlWPoRdgMu3NZtP45drfy1GKoojuR7hmRcnhZ
143145
github.com/xyproto/randomstring v1.0.5/go.mod h1:rgmS5DeNXLivK7YprL0pY+lTuhNQW3iGxZ18UQApw/E=
144146
github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
145147
github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
148+
github.com/yuin/gopher-lua v1.1.1 h1:kYKnWBjvbNP4XLT3+bPEwAXJx262OhaHDWDVOPjL46M=
149+
github.com/yuin/gopher-lua v1.1.1/go.mod h1:GBR0iDaNXjAgGg9zfCvksxSRnQx76gclCIb7kdAd1Pw=
146150
go.uber.org/automaxprocs v1.6.0 h1:O3y2/QNTOdbF+e/dpXNNW7Rx2hZ4sTIPyybbxyNqTUs=
147151
go.uber.org/automaxprocs v1.6.0/go.mod h1:ifeIMSnPZuznNm6jmdzmU3/bfk01Fe2fotchwEFJ8r8=
148152
go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=

pkg/kv-cache/kv_cache.go

Lines changed: 1 addition & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -93,10 +93,7 @@ func (h *KVCacheHelper) OnRequestStart(vllmReq openaiserverapi.CompletionRequest
9393
}
9494

9595
nBlocksAlreadyInCache, err := h.blockCache.startRequest(requestID, blockHashes)
96-
if err == nil {
97-
vllmReq.SetNumberOfCachedPromptTokens(nBlocksAlreadyInCache * h.blockSize)
98-
}
99-
96+
vllmReq.SetNumberOfCachedPromptTokens(nBlocksAlreadyInCache * h.blockSize)
10097
return err
10198
}
10299

pkg/kv-cache/kv_cache_sender.go

Lines changed: 2 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -33,11 +33,6 @@ const (
3333
eventActionRemove
3434
)
3535

36-
const (
37-
BlockStored = "BlockStored"
38-
BlockRemoved = "BlockRemoved"
39-
)
40-
4136
type EventData struct {
4237
action EventAction
4338
hashValues []uint64
@@ -98,9 +93,9 @@ func (s *KVEventSender) Run(ctx context.Context) error {
9893

9994
switch eventData.action {
10095
case eventActionStore:
101-
payload, err = msgpack.Marshal(storedToTaggedUnion(kvevents.BlockStored{BlockHashes: eventData.hashValues}))
96+
payload, err = msgpack.Marshal(kvevents.BlockStored{BlockHashes: eventData.hashValues}.ToTaggedUnion())
10297
case eventActionRemove:
103-
payload, err = msgpack.Marshal(removedToTaggedUnion(kvevents.BlockRemoved{BlockHashes: eventData.hashValues}))
98+
payload, err = msgpack.Marshal(kvevents.BlockRemoved{BlockHashes: eventData.hashValues}.ToTaggedUnion())
10499
default:
105100
return fmt.Errorf("invalid event action %d", eventData.action)
106101
}
@@ -135,24 +130,6 @@ func (s *KVEventSender) Run(ctx context.Context) error {
135130
}
136131
}
137132

138-
func storedToTaggedUnion(bs kvevents.BlockStored) []any {
139-
return []any{
140-
BlockStored,
141-
bs.BlockHashes,
142-
bs.ParentBlockHash,
143-
bs.TokenIds,
144-
bs.BlockSize,
145-
bs.LoraID,
146-
}
147-
}
148-
149-
func removedToTaggedUnion(br kvevents.BlockRemoved) []any {
150-
return []any{
151-
BlockRemoved,
152-
br.BlockHashes,
153-
}
154-
}
155-
156133
// helper to publish collected batch if not empty
157134
func (s *KVEventSender) publishHelper(ctx context.Context) error {
158135
if len(s.batch) == 0 {

pkg/kv-cache/kv_cache_test.go

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -513,11 +513,11 @@ func parseEvent(parts [][]byte, expectedTopic string, expectedSeq uint64) ([]uin
513513
Expect(err).NotTo(HaveOccurred())
514514

515515
switch tag {
516-
case BlockStored:
516+
case kvevents.BlockStoredEventTag:
517517
var bs kvevents.BlockStored
518518
err = msgpack.Unmarshal(payloadBytes, &bs)
519519
stored = append(stored, bs.BlockHashes...)
520-
case BlockRemoved:
520+
case kvevents.BlockRemovedEventTag:
521521
var br kvevents.BlockRemoved
522522
err = msgpack.Unmarshal(payloadBytes, &br)
523523
removed = append(removed, br.BlockHashes...)

0 commit comments

Comments
 (0)