Upload via push_to_hf.py
Browse files- sft/d20/meta_000700.json +23 -0
- sft/d20/model_000700.pt +3 -0
sft/d20/meta_000700.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"step": 700,
|
| 3 |
+
"val_loss": 1.078283667564392,
|
| 4 |
+
"mmlu_acc": 0.3486328125,
|
| 5 |
+
"arc_easy_acc": 0.4345703125,
|
| 6 |
+
"model_config": {
|
| 7 |
+
"sequence_len": 2048,
|
| 8 |
+
"vocab_size": 65536,
|
| 9 |
+
"n_layer": 20,
|
| 10 |
+
"n_head": 10,
|
| 11 |
+
"n_kv_head": 10,
|
| 12 |
+
"n_embd": 1280,
|
| 13 |
+
"n_prelude": 2,
|
| 14 |
+
"n_recur_block": 4,
|
| 15 |
+
"n_coda": 2,
|
| 16 |
+
"train_recur_mean": 4.0,
|
| 17 |
+
"train_recur_max": 16,
|
| 18 |
+
"recur_warm_start": true,
|
| 19 |
+
"bptt_k": 4,
|
| 20 |
+
"kv_cache_recur_budget": 1,
|
| 21 |
+
"inject_mode": "concat_linear"
|
| 22 |
+
}
|
| 23 |
+
}
|
sft/d20/model_000700.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52fa1b2df267e44297a8869c78249f6945d56e953bc2310f6c2423e4a46ba689
|
| 3 |
+
size 1145590951
|