shralex

Results 12 comments of shralex

roachtest: tpcc-nowait/isolation-level=mixed/nodes=3/w=1 failed

@arulajmani can you please take a look ?

Clarify docs on --locality-advertise-addr

@mdlinville can we prioritize this or the linked docs changes request ? I think a clarification in the docs is needed because the behavior of this flag isn't intuitive, unless...

Inconsistent environment variable names

Since this changes the "api", perhaps we could introduce duplicate variables, and mark others as deprecated while still respecting them for a while.

Flash attention - head_dim 64

@gobbleturk @RissyRan can you please take a look

Training more than one epoch

Hi Aireen, should we close this issue ?

Support Multi-Source Blending in Grain Data Pipeline

@aireenmei could you please take a look

The default setting of `param_scan_axis=1` hurts performance and memory consumption on GPUs

@khatwanimohit can we merge your PR and close this ?

a little error

@Lisennlp thank you for reporting this issue! we were curious whether you use this code ? can you provide some info about your use-case

Plans for support of Qwen models?

Hi, yes, we are looking into this. Are there particular Qwen variants you would want to see in MaxText ? Any context or motivation would be helpful.

Package hierarchy: should benchmarks, end_to_end, pedagogical_examples be under MaxText?

@bvandermoon could you please take a look