add model
Browse files- config.json +5 -5
- flax_model.msgpack +3 -0
config.json
CHANGED
|
@@ -2,15 +2,15 @@
|
|
| 2 |
"activation_dropout": 0.0,
|
| 3 |
"activation_function": "relu",
|
| 4 |
"architectures": [
|
| 5 |
-
"
|
| 6 |
],
|
| 7 |
"attention_dropout": 0.0,
|
| 8 |
"bos_token_id": 0,
|
| 9 |
-
"hidden_size": 1024,
|
| 10 |
"do_layer_norm_before": false,
|
| 11 |
"dropout": 0.1,
|
| 12 |
"eos_token_id": 2,
|
| 13 |
"ffn_dim": 4096,
|
|
|
|
| 14 |
"init_std": 0.02,
|
| 15 |
"layerdrop": 0.0,
|
| 16 |
"max_position_embeddings": 2048,
|
|
@@ -18,10 +18,10 @@
|
|
| 18 |
"num_attention_heads": 16,
|
| 19 |
"num_hidden_layers": 24,
|
| 20 |
"pad_token_id": 1,
|
|
|
|
| 21 |
"torch_dtype": "float16",
|
| 22 |
-
"transformers_version": "4.
|
| 23 |
"use_cache": true,
|
| 24 |
"vocab_size": 50272,
|
| 25 |
-
"word_embed_proj_dim": 512
|
| 26 |
-
"prefix": "</s>"
|
| 27 |
}
|
|
|
|
| 2 |
"activation_dropout": 0.0,
|
| 3 |
"activation_function": "relu",
|
| 4 |
"architectures": [
|
| 5 |
+
"OPTModel"
|
| 6 |
],
|
| 7 |
"attention_dropout": 0.0,
|
| 8 |
"bos_token_id": 0,
|
|
|
|
| 9 |
"do_layer_norm_before": false,
|
| 10 |
"dropout": 0.1,
|
| 11 |
"eos_token_id": 2,
|
| 12 |
"ffn_dim": 4096,
|
| 13 |
+
"hidden_size": 1024,
|
| 14 |
"init_std": 0.02,
|
| 15 |
"layerdrop": 0.0,
|
| 16 |
"max_position_embeddings": 2048,
|
|
|
|
| 18 |
"num_attention_heads": 16,
|
| 19 |
"num_hidden_layers": 24,
|
| 20 |
"pad_token_id": 1,
|
| 21 |
+
"prefix": "</s>",
|
| 22 |
"torch_dtype": "float16",
|
| 23 |
+
"transformers_version": "4.20.0.dev0",
|
| 24 |
"use_cache": true,
|
| 25 |
"vocab_size": 50272,
|
| 26 |
+
"word_embed_proj_dim": 512
|
|
|
|
| 27 |
}
|
flax_model.msgpack
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e70fc2225b84a806461f18321e9fc0761f1cb649801abe42a5e894b0e11ad891
|
| 3 |
+
size 662406360
|