thatdramebaazguy commited on
Commit
de652a0
1 Parent(s): 1d0c5ce

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +73 -1
README.md CHANGED
@@ -1 +1,73 @@
1
- Model Card coming soon!
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets:
3
+ - imdb (Movie corpus for Domain Adaptive Pretraining)
4
+ - cornell_movie_dialogue
5
+ - MIT Movie (NER Dataset)
6
+
7
+ language:
8
+ - English
9
+
10
+ thumbnail:
11
+
12
+ tags:
13
+ - roberta
14
+ - roberta-base
15
+ - token-classification
16
+ - NER
17
+ - named-entities
18
+ - BIO
19
+ - movies
20
+ - DAPT
21
+
22
+ license: cc-by-4.0
23
+
24
+ ---
25
+ # Movie Roberta + Movies NER Task
26
+
27
+ Objective:
28
+ This is Roberta Base + Movie DAPT --> trained for the NER task using MIT Movie Dataset
29
+ https://huggingface.co/thatdramebaazguy/movie-roberta-base was used as the MovieRoberta.
30
+
31
+ ```
32
+ model_name = "thatdramebaazguy/movie-roberta-MITmovieroberta-base-MITmovie"
33
+ pipeline(model=model_name, tokenizer=model_name, revision="v1.0", task="ner")
34
+ ```
35
+
36
+ ## Overview
37
+ **Language model:** roberta-base
38
+ **Language:** English
39
+ **Downstream-task:** NER
40
+ **Training data:** MIT Movie
41
+ **Eval data:** MIT Movie
42
+ **Infrastructure**: 2x Tesla v100
43
+ **Code:** See [example](https://github.com/adityaarunsinghal/Domain-Adaptation/blob/master/scripts/shell_scripts/movieR_NER_squad.sh)
44
+
45
+ ## Hyperparameters
46
+ ```
47
+ Num examples = 6253
48
+ Num Epochs = 5
49
+ Instantaneous batch size per device = 64
50
+ Total train batch size (w. parallel, distributed & accumulation) = 128
51
+
52
+ ```
53
+ ## Performance
54
+
55
+ ### Eval on MIT Movie
56
+ - epoch = 5.0
57
+ - eval_accuracy = 0.9472
58
+ - eval_f1 = 0.8876
59
+ - eval_loss = 0.2211
60
+ - eval_mem_cpu_alloc_delta = 3MB
61
+ - eval_mem_cpu_peaked_delta = 2MB
62
+ - eval_mem_gpu_alloc_delta = 0MB
63
+ - eval_mem_gpu_peaked_delta = 38MB
64
+ - eval_precision = 0.887
65
+ - eval_recall = 0.8881
66
+ - eval_runtime = 0:00:03.73
67
+ - eval_samples = 1955
68
+ - eval_samples_per_second = 523.095
69
+
70
+ Github Repo:
71
+ - [Domain-Adaptation Project](https://github.com/adityaarunsinghal/Domain-Adaptation/)
72
+
73
+ ---