Add dataset card

README.md

@@ -28,8 +28,8 @@ size_categories:
 Tokenized and packed pretraining data for the FORGE-3B language model.
 
 ## Stats
-- **Total tokens**: 34.1996B
-- **Domains**: 9/10
+- **Total tokens**: 36.6291B
+- **Domains**: 10/10
 - **Sequence length**: 2048 tokens
 - **Format**: `.npy` shards of shape `(N, 2048)` with dtype `uint32`
 - **Tokenizer**: CRAYON (xerv-crayon, standard profile)
@@ -43,9 +43,9 @@ Tokenized and packed pretraining data for the FORGE-3B language model.
 | wikipedia | 8% | 4.2791 | ✓ |
 | openwebmath | 8% | 3.9654 | ✓ |
 | books | 7% | 2.7678 | ✓ |
-| arxiv | 6% | 6.3538 | ✓ |
-| dolma | — | — | ✗ |
-| stackexchange | 5% | 0.0000 | ✓ |
+| arxiv | 6% | 6.6620 | ✓ |
+| dolma | 10% | 1.7980 | ✓ |
+| stackexchange | 5% | 0.3233 | ✓ |
 | redpajama_cc | 6% | 1.0675 | ✓ |
 | multilingual | 4% | 1.0200 | ✓ |