Step Final_Loss6.35_Tok19.69M
Browse files- config.json +5 -1
config.json
CHANGED
|
@@ -1,10 +1,14 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
| 2 |
"attention_dropout": 0.0,
|
| 3 |
"auto_map": {
|
| 4 |
"AutoConfig": "configuration_alinlight.AlinlightConfig",
|
| 5 |
"AutoModelForCausalLM": "modeling_alinlight.AlinlightForCausalLM"
|
| 6 |
},
|
| 7 |
"bos_token_id": 1,
|
|
|
|
| 8 |
"eos_token_id": 2,
|
| 9 |
"hidden_size": 2048,
|
| 10 |
"initializer_range": 0.02,
|
|
@@ -20,6 +24,6 @@
|
|
| 20 |
"rope_theta": 10000.0,
|
| 21 |
"sliding_window": 4096,
|
| 22 |
"transformers_version": "4.57.3",
|
| 23 |
-
"use_cache":
|
| 24 |
"vocab_size": 128000
|
| 25 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"AlinlightForCausalLM"
|
| 4 |
+
],
|
| 5 |
"attention_dropout": 0.0,
|
| 6 |
"auto_map": {
|
| 7 |
"AutoConfig": "configuration_alinlight.AlinlightConfig",
|
| 8 |
"AutoModelForCausalLM": "modeling_alinlight.AlinlightForCausalLM"
|
| 9 |
},
|
| 10 |
"bos_token_id": 1,
|
| 11 |
+
"dtype": "bfloat16",
|
| 12 |
"eos_token_id": 2,
|
| 13 |
"hidden_size": 2048,
|
| 14 |
"initializer_range": 0.02,
|
|
|
|
| 24 |
"rope_theta": 10000.0,
|
| 25 |
"sliding_window": 4096,
|
| 26 |
"transformers_version": "4.57.3",
|
| 27 |
+
"use_cache": false,
|
| 28 |
"vocab_size": 128000
|
| 29 |
}
|