Commit
·
e01e806
1
Parent(s):
bca5899
Update config.json
Browse files- config.json +8 -10
config.json
CHANGED
|
@@ -1,19 +1,18 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
"alibi": false,
|
| 4 |
"apply_residual_connection_post_layernorm": false,
|
| 5 |
"architectures": [
|
| 6 |
"RWForCausalLM"
|
| 7 |
],
|
| 8 |
"attention_dropout": 0.0,
|
| 9 |
-
"attention_probs_dropout_prob": 0.0,
|
| 10 |
"auto_map": {
|
| 11 |
-
"AutoConfig": "
|
| 12 |
-
"AutoModel": "
|
| 13 |
-
"AutoModelForCausalLM": "
|
| 14 |
-
"AutoModelForQuestionAnswering": "
|
| 15 |
-
"AutoModelForSequenceClassification": "
|
| 16 |
-
"AutoModelForTokenClassification": "
|
| 17 |
},
|
| 18 |
"bias": false,
|
| 19 |
"bos_token_id": 11,
|
|
@@ -25,7 +24,6 @@
|
|
| 25 |
},
|
| 26 |
"eos_token_id": 11,
|
| 27 |
"hidden_dropout": 0.0,
|
| 28 |
-
"hidden_dropout_prob": 0.0,
|
| 29 |
"hidden_size": 8192,
|
| 30 |
"initializer_range": 0.02,
|
| 31 |
"layer_norm_epsilon": 1e-05,
|
|
@@ -34,7 +32,7 @@
|
|
| 34 |
"n_head_kv": 8,
|
| 35 |
"n_layer": 60,
|
| 36 |
"parallel_attn": true,
|
| 37 |
-
"torch_dtype": "
|
| 38 |
"transformers_version": "4.30.0.dev0",
|
| 39 |
"use_cache": true,
|
| 40 |
"vocab_size": 65024
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "h2oai/h2ogpt-gm-oasst1-en-2048-falcon-40b-v1",
|
| 3 |
"alibi": false,
|
| 4 |
"apply_residual_connection_post_layernorm": false,
|
| 5 |
"architectures": [
|
| 6 |
"RWForCausalLM"
|
| 7 |
],
|
| 8 |
"attention_dropout": 0.0,
|
|
|
|
| 9 |
"auto_map": {
|
| 10 |
+
"AutoConfig": "configuration_RW.RWConfig",
|
| 11 |
+
"AutoModel": "modelling_RW.RWModel",
|
| 12 |
+
"AutoModelForCausalLM": "modelling_RW.RWForCausalLM",
|
| 13 |
+
"AutoModelForQuestionAnswering": "modelling_RW.RWForQuestionAnswering",
|
| 14 |
+
"AutoModelForSequenceClassification": "modelling_RW.RWForSequenceClassification",
|
| 15 |
+
"AutoModelForTokenClassification": "modelling_RW.RWForTokenClassification"
|
| 16 |
},
|
| 17 |
"bias": false,
|
| 18 |
"bos_token_id": 11,
|
|
|
|
| 24 |
},
|
| 25 |
"eos_token_id": 11,
|
| 26 |
"hidden_dropout": 0.0,
|
|
|
|
| 27 |
"hidden_size": 8192,
|
| 28 |
"initializer_range": 0.02,
|
| 29 |
"layer_norm_epsilon": 1e-05,
|
|
|
|
| 32 |
"n_head_kv": 8,
|
| 33 |
"n_layer": 60,
|
| 34 |
"parallel_attn": true,
|
| 35 |
+
"torch_dtype": "float16",
|
| 36 |
"transformers_version": "4.30.0.dev0",
|
| 37 |
"use_cache": true,
|
| 38 |
"vocab_size": 65024
|