5 Commits

Author	SHA1	Message	Date
Francis Couture-Harpin	352703b08b	test-model-random : better default tensor initialization distribution	2025-06-16 21:37:45 -04:00
Francis Couture-Harpin	dfa3c18266	tests : add LLAMA, LLAMA4, and GEMMA2 to test-model-random	2025-06-13 20:02:47 -04:00
Francis Couture-Harpin	8fe213af76	tests : avoid sprintf in test-model-random	2025-06-12 02:48:11 -04:00
Francis Couture-Harpin	7657835b33	tests : fix overflow and memory leaks in test-model-random ... * tests : fix integer types in test-model-random	2025-06-12 02:41:36 -04:00
Francis Couture-Harpin	9cd402cbe1	tests : add test-model-random ... This generates random models and then tests different concurrencies of batches to check if the output is consistent. This can detect when e.g. the recurrent cache has been broken, or anything else which would affect the consistency of the output when inferencing multiple distinct sequences. More architectures will be added, but for now this starts with Mamba. Eventually, consistency of pooled embeddings will also be tested. The goal is to reduce accidental regressions by making it easy to quickly test a lot of edge cases on the supported architectures, without having to download any model.	2025-06-12 01:00:57 -04:00