Upload 4 files
Browse files- config.json +16 -0
- model.safetensors +3 -0
- pytorch_model.bin +3 -0
- vocab.json +102 -0
config.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"d_model": 512,
|
| 3 |
+
"d_intermediate": 2048,
|
| 4 |
+
"n_layer": 8,
|
| 5 |
+
"vocab_size": 100,
|
| 6 |
+
"ssm_cfg": {
|
| 7 |
+
"layer": "Mamba2"
|
| 8 |
+
},
|
| 9 |
+
"attn_layer_idx": [],
|
| 10 |
+
"attn_cfg": {},
|
| 11 |
+
"rms_norm": true,
|
| 12 |
+
"residual_in_fp32": true,
|
| 13 |
+
"fused_add_norm": true,
|
| 14 |
+
"pad_vocab_size_multiple": 8,
|
| 15 |
+
"tie_embeddings": true
|
| 16 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:92168400e6618a7ac1d46fea8ec15aac65a96fd5af8588883097f1b20faf60c5
|
| 3 |
+
size 156162337
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38038aef6a5b1ee355e5bdf1afe1818c90422f116a012a562296cc6760ead3c2
|
| 3 |
+
size 155975700
|
vocab.json
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"C2": 78,
|
| 3 |
+
"CCCCCCCC": 81,
|
| 4 |
+
"i": 48,
|
| 5 |
+
"C1": 75,
|
| 6 |
+
"CO": 76,
|
| 7 |
+
"H]": 60,
|
| 8 |
+
"C": 26,
|
| 9 |
+
"P": 36,
|
| 10 |
+
"r": 52,
|
| 11 |
+
"CCCC": 64,
|
| 12 |
+
"]": 42,
|
| 13 |
+
"(": 6,
|
| 14 |
+
")": 7,
|
| 15 |
+
"CO)": 90,
|
| 16 |
+
"I": 30,
|
| 17 |
+
"O": 35,
|
| 18 |
+
"[": 40,
|
| 19 |
+
"c(": 74,
|
| 20 |
+
"c1": 70,
|
| 21 |
+
"@": 23,
|
| 22 |
+
"C=": 71,
|
| 23 |
+
"#": 4,
|
| 24 |
+
"H": 29,
|
| 25 |
+
"5": 17,
|
| 26 |
+
"S": 38,
|
| 27 |
+
"K": 31,
|
| 28 |
+
"[CLS]": 2,
|
| 29 |
+
"o": 51,
|
| 30 |
+
"s": 53,
|
| 31 |
+
"=": 22,
|
| 32 |
+
"C(C)": 72,
|
| 33 |
+
"(C)": 68,
|
| 34 |
+
"[C@H](": 97,
|
| 35 |
+
"M": 33,
|
| 36 |
+
"C(O)": 84,
|
| 37 |
+
"[C@@H]1": 95,
|
| 38 |
+
"[C@@H]": 66,
|
| 39 |
+
".": 10,
|
| 40 |
+
"(O": 83,
|
| 41 |
+
"6": 18,
|
| 42 |
+
"[C": 56,
|
| 43 |
+
"2": 14,
|
| 44 |
+
"[C@H]": 65,
|
| 45 |
+
"a": 43,
|
| 46 |
+
"c2cc": 91,
|
| 47 |
+
"c3": 77,
|
| 48 |
+
"e": 46,
|
| 49 |
+
"7": 19,
|
| 50 |
+
"[C@]": 82,
|
| 51 |
+
"[C@@]": 85,
|
| 52 |
+
"9": 21,
|
| 53 |
+
"[SEP]": 3,
|
| 54 |
+
"[C@@": 63,
|
| 55 |
+
"0": 12,
|
| 56 |
+
"C(=O)": 69,
|
| 57 |
+
"(O)": 67,
|
| 58 |
+
"CC(": 80,
|
| 59 |
+
"-": 9,
|
| 60 |
+
"[C@H](O)": 88,
|
| 61 |
+
"A": 24,
|
| 62 |
+
"B": 25,
|
| 63 |
+
"C=C": 89,
|
| 64 |
+
"CC": 54,
|
| 65 |
+
"[UNK]": 1,
|
| 66 |
+
"C)": 61,
|
| 67 |
+
"c2": 73,
|
| 68 |
+
"N": 34,
|
| 69 |
+
"[C@": 57,
|
| 70 |
+
"[C@@H](O)": 93,
|
| 71 |
+
"=O)": 62,
|
| 72 |
+
"O)": 55,
|
| 73 |
+
"c4": 96,
|
| 74 |
+
"L": 32,
|
| 75 |
+
"l": 49,
|
| 76 |
+
"F": 27,
|
| 77 |
+
"\\": 41,
|
| 78 |
+
"[C@H]1": 94,
|
| 79 |
+
"c(O)": 87,
|
| 80 |
+
"cc": 58,
|
| 81 |
+
"/C=C": 98,
|
| 82 |
+
"[PAD]": 0,
|
| 83 |
+
"C3": 86,
|
| 84 |
+
"%": 5,
|
| 85 |
+
"c1cc": 79,
|
| 86 |
+
"R": 37,
|
| 87 |
+
"/": 11,
|
| 88 |
+
"CC1": 99,
|
| 89 |
+
"8": 20,
|
| 90 |
+
"C(": 59,
|
| 91 |
+
"4": 16,
|
| 92 |
+
"CC(=O)": 92,
|
| 93 |
+
"1": 13,
|
| 94 |
+
"n": 50,
|
| 95 |
+
"3": 15,
|
| 96 |
+
"G": 28,
|
| 97 |
+
"c": 45,
|
| 98 |
+
"+": 8,
|
| 99 |
+
"T": 39,
|
| 100 |
+
"g": 47,
|
| 101 |
+
"b": 44
|
| 102 |
+
}
|