Commit
·
d2bb42b
1
Parent(s):
020cdb5
Push model using huggingface_hub.
Browse files- config.json +33 -2
config.json
CHANGED
|
@@ -1,7 +1,38 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"embedder": null,
|
| 3 |
-
"embedder_args":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
"empty_init": 2,
|
| 5 |
"llms": [],
|
| 6 |
-
"model_args":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"bridge_args": {
|
| 3 |
+
"bridge_type": "multi_module",
|
| 4 |
+
"hidden_dim": 2048,
|
| 5 |
+
"in_dim": 3072,
|
| 6 |
+
"out_dim": 4096
|
| 7 |
+
},
|
| 8 |
"embedder": null,
|
| 9 |
+
"embedder_args": {
|
| 10 |
+
"causal_embedder": false,
|
| 11 |
+
"compress_rates": [
|
| 12 |
+
-8
|
| 13 |
+
],
|
| 14 |
+
"cont_tok": true,
|
| 15 |
+
"n_truncated_layers": 2,
|
| 16 |
+
"pooling_module": {
|
| 17 |
+
"pool_type": "mean_pooled_queries",
|
| 18 |
+
"where": "before"
|
| 19 |
+
},
|
| 20 |
+
"rec_tok": true,
|
| 21 |
+
"train_embedding_mtx": true,
|
| 22 |
+
"trained_layers": 27
|
| 23 |
+
},
|
| 24 |
"empty_init": 2,
|
| 25 |
"llms": [],
|
| 26 |
+
"model_args": {
|
| 27 |
+
"dim": 3072,
|
| 28 |
+
"head_dim": 128,
|
| 29 |
+
"hidden_dim": 8192,
|
| 30 |
+
"max_batch_size": 1,
|
| 31 |
+
"n_heads": 24,
|
| 32 |
+
"n_kv_heads": 8,
|
| 33 |
+
"n_layers": 28,
|
| 34 |
+
"norm_eps": "1e-05",
|
| 35 |
+
"rope_theta": 500000.0,
|
| 36 |
+
"vocab_size": 128256
|
| 37 |
+
}
|
| 38 |
}
|