Hyx
Collection
Hyx model family (every HyxLabs's model) • 3 items • Updated
Trained on wikimedia/wikipedia, 20231101.en. This is bad. Device: cuda Resolving data files: 100% 41/41 [00:00<00:00, 1266.39it/s] Loaded 69940789 characters from Wikipedia Vocabulary size: 2111 Parameters: 105,039 (~1K) Step 0 | Loss: 7.6546 Step 300 | Loss: 2.6300 Step 600 | Loss: 2.6279 Step 900 | Loss: 2.6238 Step 1200 | Loss: 2.6184 Step 1500 | Loss: 2.5657 Step 1800 | Loss: 2.5590 Step 2100 | Loss: 2.5823 Step 2400 | Loss: 2.5737 Step 2700 | Loss: 2.4923 ✅ Model saved!
Sample generation: The the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the the Our first model in family. Even if this small model does only know simple English words, it is a sweetheart. ❤️💕.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support