In progress training - batch: 88
Browse files- config.json +27 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"att_dropout": 0.0,
|
| 3 |
+
"att_heads": 16,
|
| 4 |
+
"embed_dim": 1024,
|
| 5 |
+
"granular_per_slot_gate": false,
|
| 6 |
+
"head_dim": 128,
|
| 7 |
+
"interlayer_att_dropout": 0.0,
|
| 8 |
+
"interlayer_att_heads": 16,
|
| 9 |
+
"interlayer_kv_heads": 8,
|
| 10 |
+
"interlayer_qk_norm": true,
|
| 11 |
+
"kv_heads": 8,
|
| 12 |
+
"kv_input_dim": 1024,
|
| 13 |
+
"num_groups": 2,
|
| 14 |
+
"num_layers": 14,
|
| 15 |
+
"residual_gate_init": 3.0,
|
| 16 |
+
"residual_gate_type": "elementwise",
|
| 17 |
+
"rope_base": 1000000,
|
| 18 |
+
"seq_len": 8192,
|
| 19 |
+
"stm_size": 4096,
|
| 20 |
+
"use_flash_attention": false,
|
| 21 |
+
"use_gqa": true,
|
| 22 |
+
"use_granular_residual_gate": false,
|
| 23 |
+
"use_interlayer_gqa": true,
|
| 24 |
+
"use_linear_slot_status": false,
|
| 25 |
+
"use_qk_norm": true,
|
| 26 |
+
"variant": "grouped-self-interlayer"
|
| 27 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4dc3fd02ed4c3462475f605b46c21c78d579f2b73eb38a5d7cd623fa4053121
|
| 3 |
+
size 471288288
|