Codechef experiments #16
Conversation
| let mut players = std::collections::HashMap::new(); | ||
| let mut avg_perf = compute_metrics_custom(&mut players, &[]); | ||
|
|
||
| // Get list of contest names to compare with Codechef's rating system | ||
| let paths = std::fs::read_dir("/home/work_space/elommr-data/ratings").unwrap(); |
There was a problem hiding this comment.
Just to keep the organization consistent, could this be moved to "../data/codechef/old_ratings"?
(all binaries currently assume that they're being run from Elo-MMR/multi-skill)
| let mut players = std::collections::HashMap::new(); | ||
| let mut avg_perf = compute_metrics_custom(&mut players, &[]); | ||
|
|
||
| // Get list of contest names to compare with Codechef's rating system | ||
| let paths = std::fs::read_dir("/home/work_space/elommr-data/ratings").unwrap(); | ||
| let mut checkpoints = std::collections::HashSet::<String>::new(); | ||
| for path in paths { | ||
| if let Some(contest_name) = path.unwrap().path().file_stem() { |
There was a problem hiding this comment.
Is the intention to abort on all failures? If so, we can do
let contest_name = path.unwrap().path().file_stem().unwrap().to_str().unwrap();
checkpoints.insert(contest_name.to_owned());
And just to avoid the path.path(), maybe path could be named to something like file_entry.
|
|
||
| // Get list of contest names to compare with Codechef's rating system | ||
| let paths = std::fs::read_dir("/home/work_space/elommr-data/ratings").unwrap(); | ||
| let mut checkpoints = std::collections::HashSet::<String>::new(); | ||
| for path in paths { | ||
| if let Some(contest_name) = path.unwrap().path().file_stem() { | ||
| if let Some(string_name) = contest_name.to_os_string().into_string().ok() { | ||
| if let Ok(string_name) = contest_name.to_os_string().into_string() { | ||
| checkpoints.insert(string_name); | ||
| } | ||
| } | ||
| } | ||
|
|
||
| // Run the contest histories and measure | ||
| let dir = std::path::PathBuf::from("/home/work_space/elommr-data/elommr-checkpoints/codechef/"); |
There was a problem hiding this comment.
If this is just the usual contest data, shouldn't it be at cache/codechef?
| let mut players = std::collections::HashMap::new(); | ||
| let mut avg_perf = compute_metrics_custom(&mut players, &[]); | ||
|
|
||
| // Get list of contest names to compare with Codechef's rating system |
There was a problem hiding this comment.
I'd like an explanation here of what the files actually contain.
Maybe for all each of these binary files, we can clarify what the inputs and outputs are. Now would also be a good time if you want to revise the Rust file names.
| @@ -71,13 +52,12 @@ fn main() { | |||
| } | |||
|
|
|||
| // Now run the actual rating update | |||
| simulate_contest(&mut players, &contest, &*system, mu_noob, sig_noob, index); | |||
| simulate_contest(&mut players, &contest, &system, mu_noob, sig_noob, index); | |||
|
|
|||
| if checkpoints.contains(&contest.name) { | |||
| let output_file = dir.join(contest.name.clone() + ".csv"); | |||
There was a problem hiding this comment.
Output could go somewhere like
let output_dir = "..data/codechef/informative_dir_name/"
| num_contests: Option<usize>, | ||
| } | ||
|
|
||
| fn make_checkpoint(players: Vec<SimplePlayer>) -> PlayersByName { |
There was a problem hiding this comment.
Maybe include a comment that this struct & function are duplicated in another file, so I might remember to extract them someday.
| let dir = | ||
| std::path::PathBuf::from("/home/work_space/elommr-data/elommr-checkpoints/start-from-516/"); | ||
| for (index, contest) in dataset.iter().enumerate() { | ||
| if index <= contest_cutoff { |
There was a problem hiding this comment.
Replace this check with .skip(contest_cutoff + 1) on the iterator.
Btw, if you prefer indexing to start after the cutoff, you can exchange the order of .enumerate() and .skip().
| @@ -0,0 +1,122 @@ | |||
| use multi_skill::data_processing::{get_dataset_by_name, read_csv, try_write_slice_to_file}; | |||
There was a problem hiding this comment.
Some of my comments from codechef_checkpoints.rs apply to this file too, so I won't copy them here.
EbTech
force-pushed
the
master
branch
3 times, most recently
from
8a0b7d7 to
e3baa13
Compare
EbTech
force-pushed
the
master
branch
2 times, most recently
from
1f64d34 to
82c2a1c
Compare
This pull request performs a set of tests on CodeChef contest data, along with rust utilities that help parse the data we received. The notebook is withing
scripts/notebooksand can hopefully serve as a guideline for sets of tests to perform on contest data for other applications. Currently we do not have the right to release the CodeChef data, but I may make future commits to provide samples of data that the the notebook requires.