insertmike opened a new pull request, #18450: URL: https://github.com/apache/tvm/pull/18450
## Problem When loading model shards from cache (not network), the progress indicator always showed 0% because `fetchedBytes` was not incremented during the cache loading phase in `fetchTensorCacheInternal()`. The `reportCallback` function calculates progress as `fetchedBytes * 100 / totalBytes`, but `fetchedBytes` was only updated during the network download phase (line 1361), not during the cache loading phase (lines 1377-1427). This caused the progress to remain at 0% until completion when loading from cache. ## Solution This fix increments `fetchedBytes` and updates `timeElapsed` after processing each cached shard (matching the behavior of the network download phase). The progress callback now correctly reports: - Percentage completed (`fetchedBytes * 100 / totalBytes`) - MB loaded - Time elapsed ## Changes - Added `fetchedBytes += shard.nbytes;` after processing each cache shard - Added `timeElapsed` update to ensure accurate time reporting - Matches the pattern used in the download phase (lines 1360-1361) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
