sajalmandrekar commited on
Commit
19428c9
1 Parent(s): 08c7be0

uploading konkani to english model

Browse files
.gitattributes CHANGED
@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  best_model.weights.hdf5 filter=lfs diff=lfs merge=lfs -text
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  best_model.weights.hdf5 filter=lfs diff=lfs merge=lfs -text
37
+ tbase_kok-en.hdf5 filter=lfs diff=lfs merge=lfs -text
tbase_en-kok_config.env ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # -----Configurations of the Transformer model----- #
2
+
3
+ # Model name
4
+ MODEL_NAME=TRANS_BASE_EK
5
+
6
+ ## Path to training data of source language
7
+ CONTEXT_DATA_PATH=dataset/FULL_DATA.en
8
+
9
+ ## Path to training data of target language
10
+ TARGET_DATA_PATH=dataset/FULL_DATA.gom
11
+
12
+ ## Path to vocabulary of source language
13
+ CONTEXT_TOKEN_PATH=vocabulary/bert_en.vocab
14
+
15
+ ## Path to vocabulary data of target language
16
+ TARGET_TOKEN_PATH=vocabulary/bert_gom.vocab
17
+
18
+ # Reloading weights from pretrained model (Comment out or leave empty or set to 'None' if not using)
19
+ WEIGHTS_PATH=trained_models/T_BASE_EK_07_07/checkpoints/best_model.weights.hdf5
20
+
21
+ # Set model configurations
22
+
23
+ BATCH_SIZE=128
24
+ MAX_TOKENS=128
25
+
26
+ ## number of encoder and decoder layers
27
+ NUM_LAYERS=6
28
+
29
+ ## dimensionality of the embeddings
30
+ D_MODEL=512
31
+
32
+ ## internal dimensionality of the FeedForward layer
33
+ DFF=2048
34
+
35
+ ## The number of self-attention heads
36
+ NUM_HEADS=8
37
+
38
+ ## Residual Dropout
39
+ DROPOUT_RATE=0.3
40
+
41
+
42
+ # Set Training parameters
43
+ epochs=10
44
+
45
+ ## save only the best weight? default = True
46
+ save_best_only=True
47
+
48
+ ## save epochs after every `save_freq`, default = 1
49
+ save_freq=1
tbase_kok-en.hdf5 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb4669cdc8d3ec5ce86f3e843e4f1dcf8925ede6e51894f6111b8d2836a304a8
3
+ size 1231670576
tbase_kok-en_config.env ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # -----Configurations of the Transformer model----- #
2
+
3
+ # Model name
4
+ MODEL_NAME=TRANS_BASE_KE
5
+
6
+ ## Path to training data of source language
7
+ CONTEXT_DATA_PATH=dataset/FULL_DATA.gom
8
+
9
+ ## Path to training data of target language
10
+ TARGET_DATA_PATH=dataset/FULL_DATA.en
11
+
12
+ ## Path to vocabulary of source language
13
+ CONTEXT_TOKEN_PATH=vocabulary/bert_gom.vocab
14
+
15
+ ## Path to vocabulary data of target language
16
+ TARGET_TOKEN_PATH=vocabulary/bert_en.vocab
17
+
18
+ # Reloading weights from pretrained model (Comment out or leave empty or set to 'None' if not using)
19
+ WEIGHTS_PATH=trained_models/T_BASE_KE_17_07/checkpoints/best_model.weights.hdf5
20
+
21
+ # Set model configurations
22
+
23
+ BATCH_SIZE=128
24
+ MAX_TOKENS=128
25
+
26
+ ## number of encoder and decoder layers
27
+ NUM_LAYERS=6
28
+
29
+ ## dimensionality of the embeddings
30
+ D_MODEL=512
31
+
32
+ ## internal dimensionality of the FeedForward layer
33
+ DFF=2048
34
+
35
+ ## The number of self-attention heads
36
+ NUM_HEADS=8
37
+
38
+ ## Residual Dropout
39
+ DROPOUT_RATE=0.3
40
+
41
+
42
+ # Set Training parameters
43
+ epochs=10
44
+
45
+ ## save only the best weight? default = True
46
+ save_best_only=True
47
+
48
+ ## save epochs after every `save_freq`, default = 1
49
+ save_freq=1