Update README.md
Browse files
README.md
CHANGED
|
@@ -1,20 +1,44 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
tags:
|
| 4 |
-
-
|
|
|
|
| 5 |
metrics:
|
| 6 |
- bleu
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
model-index:
|
| 8 |
-
- name:
|
| 9 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
---
|
| 11 |
|
| 12 |
-
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 13 |
-
should probably proofread and complete it, then remove this comment. -->
|
| 14 |
-
|
| 15 |
# h3
|
| 16 |
|
| 17 |
-
This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on
|
|
|
|
| 18 |
It achieves the following results on the evaluation set:
|
| 19 |
- Loss: 0.2782
|
| 20 |
- Exact Match: 0.2879
|
|
@@ -28,15 +52,13 @@ It achieves the following results on the evaluation set:
|
|
| 28 |
|
| 29 |
## Model description
|
| 30 |
|
| 31 |
-
|
|
|
|
|
|
|
| 32 |
|
| 33 |
## Intended uses & limitations
|
| 34 |
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
## Training and evaluation data
|
| 38 |
-
|
| 39 |
-
More information needed
|
| 40 |
|
| 41 |
## Training procedure
|
| 42 |
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
tags:
|
| 4 |
+
- distigpt2
|
| 5 |
+
- hearthstone
|
| 6 |
metrics:
|
| 7 |
- bleu
|
| 8 |
+
- dvitel/codebleu
|
| 9 |
+
- exact_match
|
| 10 |
+
- chrf
|
| 11 |
+
datasets:
|
| 12 |
+
- dvitel/hearthstone
|
| 13 |
model-index:
|
| 14 |
+
- name: h0
|
| 15 |
+
results:
|
| 16 |
+
- task:
|
| 17 |
+
type: text-generation
|
| 18 |
+
name: Python Code Synthesis
|
| 19 |
+
dataset:
|
| 20 |
+
type: dvitel/hearthstone
|
| 21 |
+
name: HearthStone
|
| 22 |
+
split: test
|
| 23 |
+
metrics:
|
| 24 |
+
- type: exact_match
|
| 25 |
+
value: 0.30303030303030304
|
| 26 |
+
name: Exact Match
|
| 27 |
+
- type: bleu
|
| 28 |
+
value: 0.8850182403024257
|
| 29 |
+
name: BLEU
|
| 30 |
+
- type: dvitel/codebleu
|
| 31 |
+
value: 0.677852377992836
|
| 32 |
+
name: CodeBLEU
|
| 33 |
+
- type: chrf
|
| 34 |
+
value: 91.00848749530383
|
| 35 |
+
name: chrF
|
| 36 |
---
|
| 37 |
|
|
|
|
|
|
|
|
|
|
| 38 |
# h3
|
| 39 |
|
| 40 |
+
This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on [hearthstone](https://huggingface.co/datasets/dvitel/hearthstone) dataset.
|
| 41 |
+
[GitHub repo](https://github.com/dvitel/nlp-sem-parsing/blob/master/h3.py).
|
| 42 |
It achieves the following results on the evaluation set:
|
| 43 |
- Loss: 0.2782
|
| 44 |
- Exact Match: 0.2879
|
|
|
|
| 52 |
|
| 53 |
## Model description
|
| 54 |
|
| 55 |
+
DistilGPT2 fine-tuned on HearthStone dataset for 200 epochs. \
|
| 56 |
+
Related to [dvitel/h0](https://huggingface.co/dvitel/h0) but with preprocessing which anonymizes classes and function variables (Local renaming). \
|
| 57 |
+
[dvitel/h2](https://huggingface.co/dvitel/h2) implements global renaming where all names are removed. Global renaming showed worse results compared to local renaming.
|
| 58 |
|
| 59 |
## Intended uses & limitations
|
| 60 |
|
| 61 |
+
HearthStone card code synthesis.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
|
| 63 |
## Training procedure
|
| 64 |
|