nsaroiu commited on
Commit
9bb66a6
·
verified ·
1 Parent(s): 1fd62ed

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +10 -9
README.md CHANGED
@@ -15,7 +15,7 @@ datasets:
15
  # MTG Commander Deck Completion
16
 
17
  Given a Commander (and optionally a partial decklist), suggests which other
18
- cards belong in the deck. Trained on 122,571 real public Commander decklists
19
  from [Moxfield](https://moxfield.com) ([dataset](https://huggingface.co/datasets/nsaroiu/moxfield-dump)).
20
 
21
  Training/evaluation code: [github.com/nsaroiu/mtg-deck-completion](https://github.com/nsaroiu/mtg-deck-completion)
@@ -74,13 +74,13 @@ empty), staple-recognition checkpoint, validation split:
74
 
75
  | mask ratio | Recall@50 | Precision@50 | MRR |
76
  |---|---|---|---|
77
- | 0.1 (deck mostly complete) | 0.60 | 0.10 | 0.132 |
78
- | 0.5 | 0.51 | 0.44 | 0.068 |
79
- | 0.9 (deck mostly empty) | 0.38 | 0.59 | 0.046 |
80
 
81
  Held-out-commander split (commanders never seen in training, testing
82
  generalization via card content rather than memorized co-occurrence):
83
- Recall@50 0.27–0.41 across the same ratio range — meaningfully above chance,
84
  confirming the hybrid content embeddings carry real signal for unfamiliar
85
  commanders. Full breakdown by card-popularity tier, plus the synergy/pruning
86
  checkpoints' own numbers, in the GitHub repo's eval output.
@@ -151,7 +151,8 @@ fusion, if that's all you need.
151
  ## License
152
 
153
  The weights and code in this repository are released under the MIT license.
154
- The training data itself is a scrape of public Moxfield decklists with no
155
- asserted license (see the [dataset card](https://huggingface.co/datasets/nsaroiu/moxfield-dump))
156
- — if that matters for your use case, that's Moxfield's terms to check, not
157
- a constraint this repo imposes.
 
 
15
  # MTG Commander Deck Completion
16
 
17
  Given a Commander (and optionally a partial decklist), suggests which other
18
+ cards belong in the deck. Trained on 189,108 real public Commander decklists
19
  from [Moxfield](https://moxfield.com) ([dataset](https://huggingface.co/datasets/nsaroiu/moxfield-dump)).
20
 
21
  Training/evaluation code: [github.com/nsaroiu/mtg-deck-completion](https://github.com/nsaroiu/mtg-deck-completion)
 
74
 
75
  | mask ratio | Recall@50 | Precision@50 | MRR |
76
  |---|---|---|---|
77
+ | 0.1 (deck mostly complete) | 0.57 | 0.10 | 0.121 |
78
+ | 0.5 | 0.48 | 0.40 | 0.066 |
79
+ | 0.9 (deck mostly empty) | 0.36 | 0.54 | 0.045 |
80
 
81
  Held-out-commander split (commanders never seen in training, testing
82
  generalization via card content rather than memorized co-occurrence):
83
+ Recall@50 0.29–0.44 across the same ratio range — meaningfully above chance,
84
  confirming the hybrid content embeddings carry real signal for unfamiliar
85
  commanders. Full breakdown by card-popularity tier, plus the synergy/pruning
86
  checkpoints' own numbers, in the GitHub repo's eval output.
 
151
  ## License
152
 
153
  The weights and code in this repository are released under the MIT license.
154
+ The training data itself is a scrape of public Moxfield decklists (see the
155
+ [dataset repo](https://huggingface.co/datasets/nsaroiu/moxfield-dump), now
156
+ gated/manually-approved) with no asserted license — if that matters for
157
+ your use case, that's Moxfield's terms to check, not a constraint this repo
158
+ imposes.