Andrew Fitzgibbon

Results 30 comments of Andrew Fitzgibbon

My proposal is to change all to row-major. But that actually disfavours Julia, which is col-major by default... More thought needed.

2. replace pack() unpack() with struct as above

3. Replace ``parents`` integer indices with a tree data structure. This will allow for less mutative updating in ``relatives_to_absolutes`` and cleaner, hopefully faster, code.

Oddly it's 2>file, e.g. `` & {write-host "una"; "conor"; write-error "fed"} 2>err.txt >out.txt ``

The documentation [here](https://docs.wandb.ai/guides/data-vis/log-tables#add-data) also caused me to believe that `add_data` did something. And here https://docs.wandb.ai/guides/data-vis/log-tables#log-a-table-to-a-run Please do fix the doc if the actual fix is going to take time.

The specific use-case of interest to me is to log ~100 numbers (layer statistics, say) per minibatch. I would expect to add one row per iteration.

> The documentation [here](https://docs.wandb.ai/guides/data-vis/log-tables#add-data) also caused me to believe that `add_data` did something. > > And here https://docs.wandb.ai/guides/data-vis/log-tables#log-a-table-to-a-run > > Please do fix the doc if the actual fix is...

No, good to have 1.0 only.

@nalimilan Yes, pulling the fix from that thread back to here as it is currently an issue in JSON parsefile that means that if I call parsefile with default arguments,...