mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-23 23:18:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

kvcache: Pass granular cache size into implementations

Browse Source 
Currently the runner computes the kv size needed and creates a
cache of that size. This is the context size times number of
parallel sequences.

Cache implementations can make better decisions about their memory
usage, so instead pass in the required capacity, number of sequences
and maximum batch size. For now, the causal cache just uses this to
compute the size in the same way as before.
This commit is contained in:
Jesse Gross
2025-03-18 14:31:52 -07:00
committed by Jesse Gross
parent 6d1103048e
commit 3ed7ad3ab3
7 changed files with 44 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
kvcache/encoder.go
View File

@@ -49,7 +49,7 @@ func NewEncoderCache() *EncoderCache {
}
}
func (c *EncoderCache) Init(backend ml.Backend, dtype ml.DType, capacity int32) {
func (c *EncoderCache) Init(backend ml.Backend, dtype ml.DType, maxSequences, capacity, maxBatch int) {
if c.config == nil {
var config ml.CacheConfig
if cc, ok := backend.(ml.BackendCacheConfig); ok {
@@ -58,6 +58,10 @@ func (c *EncoderCache) Init(backend ml.Backend, dtype ml.DType, capacity int32)
c.config = &config
}
if maxSequences > 1 {
panic(fmt.Errorf("encoder cache does not support multiple sequences; requested: %v", maxSequences))
}
if c.config.CachePadding != 0 && c.config.CachePadding != 1 {
panic(fmt.Errorf("encoder cache is unable to enforce requested CachePadding (%v)", c.config.CachePadding))
}