Fikavec

@ldevyataykina read https://huggingface.co/blog/how-to-generate and try change num_beams, no_repeat_ngram_size and other parameters from article. > translation_pipeline = pipeline( > "translation", > model=model, > tokenizer=tokenizer, > src_lang="rus_Cyrl", > tgt_lang="eng_Latn", > max_length=512, >...

> Hi @Fikavec , Why did you close this issue? Hi @alonre24, I tested same workload on [redis-stack-server/latest](https://hub.docker.com/layers/redis/redis-stack-server/latest/images/sha256-e13c8531d1c16b58b822182933866013d4d32353a7790557f8972132e8555d1b?context=explore) image with RediSearch 2.8.11 and bug still here, but without messages in...

> When I try to reproduce, after loading 1M vectors the memory consumption by redis is ~3.3GB (not 5-6). Does your schema consist of `FLAT` vector field only as seen...

Thanks, it became clear why it was not possible to create an index with a large INITIAL_CAP (in my case, there are a lot of one search+one indexing operations, this...

> Vectors in Redis are stored twice today Is it also true for vectors stored as fields in JSON documents? Is there an option to not store vectors at all...

The previously attached script [redis_hang_test.zip](https://github.com/RediSearch/RediSearch/files/14321766/redis_hang_test.zip) also caused a connection hang on server. It can be used to reproduce and test the problem.

Still hangs. I tested same real and test workload on docker redis-stack-server/latest image with Redis 7.2.4 RediSearch 2.8.12 (Git=2.8-32fdaca).

16 days of round-the-clock RediSearch 2.8.13 (Git=2.8-c1535c6) testing - no problems, the bug has been fixed, thanks!