For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python class

VisionPreprocessCacheStats

VisionPreprocessCacheStats

class max.pipelines.modeling.types.VisionPreprocessCacheStats(hits, misses, evictions, size_bytes, capacity_bytes)

source

Bases: object

One media preprocess cache’s counters, as of now.

The three counts are cumulative for the lifetime of the process that owns the cache, and the two byte figures are instantaneous. A reader that wants rates owns the differencing, and has to survive the counts restarting at zero: the cache is rebuilt empty in any process it is unpickled into.

Parameters:

• hits (int)
• misses (int)
• evictions (int)
• size_bytes (int)
• capacity_bytes (int)

capacity_bytes

capacity_bytes: int

source

The byte budget, which size_bytes is evicted down to (0: caching is disabled).

evictions

evictions: int

source

Entries dropped to stay inside the byte budget.

hits

hits: int

source

Lookups the cache answered without preprocessing.

misses

misses: int

source

Lookups that had to preprocess.

size_bytes

size_bytes: int

source

Host bytes the cached payloads retain right now.