Upload folder using huggingface_hub

args.yaml

@@ -51,7 +51,7 @@ local_rank: null
 loglevel: INFO
 loglevel_secondary_workers: INFO
 max_checkpoints: null
-max_num_checkpoint_not_improved: 50
+max_num_checkpoint_not_improved: 100
 max_num_epochs: null
 max_samples: null
 max_seconds: null
@@ -59,7 +59,7 @@ max_seq_len:
 - 512
 - 512
 max_updates: null
-min_num_epochs: null
+min_num_epochs: 10
 min_samples: null
 min_updates: null
 momentum: 0.0
@@ -83,11 +83,11 @@ optimizer_betas:
 - 0.9
 - 0.999
 optimizer_eps: 1.0e-08
-output: /shares/volk.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-tuned/model
+output: /shares/iict-sp2.ebling.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-gpt-tuned/model
 overwrite_output: false
 pad_vocab_to_multiple_of: 8
-params: /shares/volk.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-v4/model/params.best
-prepared_data: /shares/volk.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-tuned/train_data
+params: /shares/iict-sp2.ebling.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-gpt/model/params.best
+prepared_data: /shares/iict-sp2.ebling.cl.uzh/amoryo/checkpoints/signwriting-translation/spoken-to-signed/target-factors-gpt-tuned/train_data
 quiet: false
 quiet_secondary_workers: false
 seed: 1

config

@@ -2,30 +2,30 @@
 config_data: !DataConfig
   data_statistics: !DataStatistics
     average_len_target_per_bucket:
-    - 6.01988730121871
-    - 10.49707202236441
-    - 18.536005711954232
-    - 25.91453940066591
-    - 33.44474802671516
-    - 41.29017292281736
-    - 50.38165137614682
-    - 59.57928802588991
-    - 67.86491420226359
-    - 75.76032388663967
-    - 82.68249075215772
-    - 12.0
-    - 23.333333333333332
-    - 24.88888888888889
-    - 15.857142857142856
-    - 8.6
-    - 8.0
-    - 6.5
-    - 4.5
-    - null
-    - 6.25
+    - 6.011277314861252
+    - 10.39671010650862
+    - 18.504618481112797
+    - 26.041977712863865
+    - 32.48293857888395
+    - 41.60258525852575
+    - 51.3669121514993
+    - 59.76513060097686
+    - 68.25156398104271
+    - 75.67976141505562
+    - 83.09476309226905
+    - 21.1875
+    - 7.749999999999999
+    - 5.833333333333333
     - 6.0
-    - 8.0
-    - 10.0
+    - 7.0
+    - 7.25
+    - 6.333333333333333
+    - 7.0
+    - 5.0
+    - null
+    - null
+    - null
+    - null
     - null
     - null
     - null
@@ -263,80 +263,73 @@ config_data: !DataConfig
     - !!python/tuple
       - 513
       - 513
-    length_ratio_mean: 1.4422861755889766
+    length_ratio_mean: 1.4462363190719616
     length_ratio_stats_per_bucket:
     - !!python/tuple
-      - 1.1255982284370691
-      - 0.3093025169709561
-    - !!python/tuple
-      - 1.6064942041128283
-      - 0.6417946634210556
-    - !!python/tuple
-      - 2.359196679910085
-      - 0.9575904838242857
+      - 1.13785261776882
+      - 0.30617192612461125
     - !!python/tuple
-      - 2.9607196399710274
-      - 1.5584344652106463
+      - 1.6532001104346803
+      - 0.6379238313847178
     - !!python/tuple
-      - 2.9543389770526325
-      - 2.297144978089179
+      - 2.471066932954014
+      - 0.9542933578872858
     - !!python/tuple
-      - 3.002340450312629
-      - 2.18681603522054
+      - 2.9657919123449297
+      - 1.5793440289847793
     - !!python/tuple
-      - 3.081058473996043
-      - 2.5011199006515183
+      - 3.096168086847911
+      - 2.441325976657657
     - !!python/tuple
-      - 2.9163894753348045
-      - 2.2789742432670086
+      - 3.268248874962792
+      - 2.2403195790511785
     - !!python/tuple
-      - 2.9722600022754504
-      - 2.230352263818367
+      - 3.453956732311208
+      - 2.518686961607376
     - !!python/tuple
-      - 3.138369081443168
-      - 2.4620594208603235
+      - 3.2456624469723585
+      - 2.2619507905455962
     - !!python/tuple
-      - 2.854253034817575
-      - 1.6376340581023332
+      - 3.322785538319874
+      - 2.1624153548695157
     - !!python/tuple
-      - 0.12974460739238317
-      - 0.160679251677897
+      - 3.489342008082259
+      - 2.469578333618206
     - !!python/tuple
-      - 0.23436597326420017
-      - 0.24015947073924906
+      - 3.278259980416589
+      - 1.6906831932492183
     - !!python/tuple
-      - 0.23457181630569815
-      - 0.2675720247460546
+      - 0.23045117210041233
+      - 0.26501940303452104
     - !!python/tuple
-      - 0.1366625632624212
-      - 0.16557849394412172
+      - 0.07818400556098729
+      - 0.027479091254378488
     - !!python/tuple
-      - 0.06941451000342352
-      - 0.02389100863634256
+      - 0.05445908159715249
+      - 0.013332869808672716
     - !!python/tuple
-      - 0.05970149253731343
+      - 0.05042016806722689
       - 0.0
     - !!python/tuple
-      - 0.04609929078014184
-      - 0.017730496453900707
+      - 0.05658536585365853
+      - 0.016585365853658534
     - !!python/tuple
-      - 0.03063680118474639
-      - 0.0036097741577193615
-    - &id001 !!python/tuple
-      - null
-      - null
+      - 0.05537665840439907
+      - 0.021182928316626336
     - !!python/tuple
-      - 0.038657633805228714
-      - 0.010237907973024957
+      - 0.04568795432892477
+      - 0.009015856562239487
     - !!python/tuple
-      - 0.03469146238377008
-      - 0.004700847851307781
-    - !!python/tuple
-      - 0.0446927374301676
+      - 0.047619047619047616
       - 0.0
     - !!python/tuple
-      - 0.05405405405405406
+      - 0.032679738562091505
       - 0.0
+    - &id001 !!python/tuple
+      - null
+      - null
+    - *id001
+    - *id001
     - *id001
     - *id001
     - *id001
@@ -378,36 +371,37 @@ config_data: !DataConfig
     - *id001
     - *id001
     - *id001
-    length_ratio_std: 0.8479651722336846
-    max_observed_len_source: 185
+    - *id001
+    length_ratio_std: 0.8476478699398606
+    max_observed_len_source: 153
     max_observed_len_target: 87
     num_discarded: 0
-    num_sents: 425392
+    num_sents: 779247
     num_sents_per_bucket:
-    - 245433
-    - 135930
-    - 19608
-    - 7208
-    - 3294
-    - 2371
-    - 2180
-    - 2472
-    - 2739
-    - 2470
-    - 1622
-    - 20
-    - 9
-    - 9
-    - 7
-    - 5
-    - 1
-    - 2
+    - 478128
+    - 233503
+    - 27823
+    - 9243
+    - 4982
+    - 3636
+    - 3802
+    - 4709
+    - 5275
+    - 4862
+    - 3208
+    - 32
+    - 8
+    - 12
     - 2
-    - 0
     - 4
-    - 3
-    - 1
+    - 8
+    - 6
     - 2
+    - 2
+    - 0
+    - 0
+    - 0
+    - 0
     - 0
     - 0
     - 0
@@ -449,11 +443,11 @@ config_data: !DataConfig
     - 0
     - 0
     - 0
-    num_tokens_source: 3056735
-    num_tokens_target: 4427820
+    num_tokens_source: 5305413
+    num_tokens_target: 7842731
     num_unks_source: 0
     num_unks_target: 0
-    size_vocab_source: 8016
+    size_vocab_source: 45864
     size_vocab_target: 656
   eop_id: -1
   max_seq_len_source: 513
@@ -485,7 +479,7 @@ config_embed_source: !EmbeddingConfig
   factor_configs: null
   num_embed: 512
   num_factors: 1
-  vocab_size: 8016
+  vocab_size: 45864
 config_embed_target: !EmbeddingConfig
   allow_sparse_grad: false
   dropout: 0.5
@@ -537,6 +531,6 @@ dtype: float32
 lhuc: false
 neural_vocab_selection: null
 neural_vocab_selection_block_loss: false
-vocab_source_size: 8016
+vocab_source_size: 45864
 vocab_target_size: 656
 weight_tying_type: trg_softmax

params.best

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fbc40d947ce0a1c03679cc4c62a4a94c11b179a485637727fc1b491a063042c4
-size 201248510
+oid sha256:c3cbabac4d28a48fbef952e0155e471556d607dda6700a919f08b44b49299597
+size 278761214