Improve the 3-class sentence-based model.
Browse filesIn the previous training we inadvertently used a sequence length of just `20` tokens,
which is what `AutoConfig.from_pretrained defaults to. This model has been trained
with up to 512 tokens, better utilizing Roberta's capabilities.
F1 macro score is now `0.6951`.
- config.json +2 -1
- pytorch_model.bin +1 -1
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "./artifacts/model-
|
3 |
"architectures": [
|
4 |
"RobertaForSequenceClassification"
|
5 |
],
|
@@ -24,6 +24,7 @@
|
|
24 |
"intermediate": 1
|
25 |
},
|
26 |
"layer_norm_eps": 1e-05,
|
|
|
27 |
"max_position_embeddings": 514,
|
28 |
"model_type": "roberta",
|
29 |
"num_attention_heads": 12,
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "./artifacts/model-1qe3kbqj-2742-0.6951:v0",
|
3 |
"architectures": [
|
4 |
"RobertaForSequenceClassification"
|
5 |
],
|
|
|
24 |
"intermediate": 1
|
25 |
},
|
26 |
"layer_norm_eps": 1e-05,
|
27 |
+
"max_length": 512,
|
28 |
"max_position_embeddings": 514,
|
29 |
"model_type": "roberta",
|
30 |
"num_attention_heads": 12,
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 498675309
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:26854ee4fa633135cd6bc6f190398ae0cca099a8d3e9adeb54c94bb61a536530
|
3 |
size 498675309
|