Upload 3 files
Browse files
convert/sg/mixtral-8x7b-32kseqlen/config.json
ADDED
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"dim": 4096,
|
3 |
+
"hidden_dim": 14336,
|
4 |
+
"head_dim": 128,
|
5 |
+
"n_layers": 32,
|
6 |
+
"n_heads": 32,
|
7 |
+
"n_kv_heads": 8,
|
8 |
+
"vocab_size": 32000,
|
9 |
+
"norm_eps": 1e-05,
|
10 |
+
"rope_theta": 1000000,
|
11 |
+
"max_batch_size": 32,
|
12 |
+
"max_seq_len": 4096,
|
13 |
+
"moe": {
|
14 |
+
"num_experts_per_tok": 2,
|
15 |
+
"num_experts": 8
|
16 |
+
},
|
17 |
+
"rope_scaling": null
|
18 |
+
}
|
convert/sg/mixtral-8x7b-32kseqlen/meta.json
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"llama_type": "mistral"
|
3 |
+
}
|
convert/sg/mixtral-8x7b-32kseqlen/tokenizer.model
ADDED
Binary file (493 kB). View file
|
|