Commit
•
833b100
1
Parent(s):
d0e2be6
update model card README.md
Browse files
README.md
CHANGED
@@ -1,28 +1,25 @@
|
|
1 |
---
|
2 |
-
license:
|
3 |
-
|
4 |
-
|
5 |
-
-
|
6 |
-
-
|
7 |
-
-
|
8 |
-
|
9 |
-
- text: "My name is Thomas and my main"
|
10 |
-
inference:
|
11 |
-
parameters:
|
12 |
-
max_length: 200
|
13 |
---
|
14 |
|
15 |
-
|
16 |
-
|
17 |
|
18 |
-
#
|
19 |
-
The model architecture is based on the GPT-2 framework, specifically using the parameters of the small version of the original OpenAI GPT2 model. It employs a Byte Pair Encoding (BPE) tokenizer.
|
20 |
|
21 |
-
|
22 |
-
|
|
|
23 |
|
24 |
-
|
25 |
-
|
|
|
26 |
|
27 |
## Intended uses & limitations
|
28 |
|
@@ -34,38 +31,34 @@ More information needed
|
|
34 |
|
35 |
## Training procedure
|
36 |
|
37 |
-
More information needed
|
38 |
-
|
39 |
### Training hyperparameters
|
40 |
|
41 |
The following hyperparameters were used during training:
|
42 |
- learning_rate: 0.0005
|
43 |
-
- train_batch_size:
|
44 |
-
- eval_batch_size:
|
45 |
- seed: 42
|
46 |
- gradient_accumulation_steps: 4
|
47 |
-
- total_train_batch_size:
|
48 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
49 |
- lr_scheduler_type: cosine
|
50 |
-
- lr_scheduler_warmup_steps:
|
51 |
- num_epochs: 1
|
52 |
-
- mixed_precision_training: Native AMP
|
53 |
|
54 |
### Training results
|
55 |
|
56 |
-
|
|
57 |
-
|
58 |
-
|
|
59 |
-
|
|
60 |
-
|
|
61 |
-
|
|
62 |
-
|
|
63 |
-
| 3000 | 0.8852 | 0.863592 |
|
64 |
|
65 |
|
66 |
### Framework versions
|
67 |
|
68 |
-
- Transformers 4.
|
69 |
-
-
|
70 |
-
- Datasets 2.
|
71 |
- Tokenizers 0.13.3
|
|
|
1 |
---
|
2 |
+
license: mit
|
3 |
+
base_model: gpt2
|
4 |
+
tags:
|
5 |
+
- generated_from_trainer
|
6 |
+
model-index:
|
7 |
+
- name: hindi_gpt2
|
8 |
+
results: []
|
|
|
|
|
|
|
|
|
9 |
---
|
10 |
|
11 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
12 |
+
should probably proofread and complete it, then remove this comment. -->
|
13 |
|
14 |
+
# hindi_gpt2
|
|
|
15 |
|
16 |
+
This model is a fine-tuned version of [gpt2](https://huggingface.co/gpt2) on the None dataset.
|
17 |
+
It achieves the following results on the evaluation set:
|
18 |
+
- Loss: 1.9187
|
19 |
|
20 |
+
## Model description
|
21 |
+
|
22 |
+
More information needed
|
23 |
|
24 |
## Intended uses & limitations
|
25 |
|
|
|
31 |
|
32 |
## Training procedure
|
33 |
|
|
|
|
|
34 |
### Training hyperparameters
|
35 |
|
36 |
The following hyperparameters were used during training:
|
37 |
- learning_rate: 0.0005
|
38 |
+
- train_batch_size: 40
|
39 |
+
- eval_batch_size: 40
|
40 |
- seed: 42
|
41 |
- gradient_accumulation_steps: 4
|
42 |
+
- total_train_batch_size: 160
|
43 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
44 |
- lr_scheduler_type: cosine
|
45 |
+
- lr_scheduler_warmup_steps: 400
|
46 |
- num_epochs: 1
|
|
|
47 |
|
48 |
### Training results
|
49 |
|
50 |
+
| Training Loss | Epoch | Step | Validation Loss |
|
51 |
+
|:-------------:|:-----:|:----:|:---------------:|
|
52 |
+
| 4.694 | 0.18 | 400 | 2.7361 |
|
53 |
+
| 2.3952 | 0.35 | 800 | 2.1608 |
|
54 |
+
| 2.1311 | 0.53 | 1200 | 2.0237 |
|
55 |
+
| 2.0282 | 0.71 | 1600 | 1.9518 |
|
56 |
+
| 1.9731 | 0.89 | 2000 | 1.9187 |
|
|
|
57 |
|
58 |
|
59 |
### Framework versions
|
60 |
|
61 |
+
- Transformers 4.31.0
|
62 |
+
- Pytorch 2.0.1+cu118
|
63 |
+
- Datasets 2.14.2
|
64 |
- Tokenizers 0.13.3
|