Upload folder using huggingface_hub

best_model.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:83b8bf7c5192c5c6b868bf38fc1852d06524e8e75a507b623500dfe613e5e95d
-size 100239739
+oid sha256:d046b535626fd43d2e906f1d7810429dead000ffafd5889156cb9c1df1cd39ee
+size 124262907

config.json

@@ -1,9 +1,4 @@
 {
-  "run_name": "c3",
-  "positional_encoding": "sinusoidal",
-  "attention_type": "gqa",
-  "norm_type": "layernorm",
-  "tokenization": "subword",
   "d_model": 256,
   "num_heads": 8,
   "num_encoder_layers": 4,
@@ -14,10 +9,21 @@
   "learning_rate": 0.0003,
   "warmup_steps": 2000,
   "batch_size": 64,
+  "grad_clip_norm": 1.0,
   "max_seq_len": 512,
   "seed": 42,
   "epochs": 50,
-  "num_params": 8291648,
-  "device": "cuda",
-  "gqa_kv_heads": 2
+  "patience": 5,
+  "bpe_vocab_size": 8000,
+  "positional_encoding": "sinusoidal",
+  "attention_type": "gqa",
+  "norm_type": "layernorm",
+  "tokenization": "subword",
+  "gqa_kv_heads": 2,
+  "blt_patch_size": 9,
+  "d_local": 256,
+  "num_local_layers": 4,
+  "run_name": "c3",
+  "num_params": 10293568,
+  "device": "cuda"
 }