End of training

README.md

@@ -1,10 +1,9 @@
 ---
+base_model: Lakoc/DeCRED_small_cv
 tags:
 - generated_from_trainer
 datasets:
 - fleurs
-metrics:
-- wer
 model-index:
 - name: experiment
   results: []
@@ -15,18 +14,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # experiment
 
-This model is a fine-tuned version of [](https://huggingface.co/) on the fleurs dataset.
-It achieves the following results on the evaluation set:
-- Loss: 7.9280
-- Cer: 1.0
-- Wer: 1.0
-- Mer: 1.0
-- Wil: 1
-- Wip: 0
-- Hits: 0
-- Substitutions: 0
-- Deletions: 34
-- Insertions: 0
+This model is a fine-tuned version of [Lakoc/DeCRED_small_cv](https://huggingface.co/Lakoc/DeCRED_small_cv) on the fleurs dataset.
 
 ## Model description
 
@@ -53,20 +41,9 @@ The following hyperparameters were used during training:
 - lr_scheduler_type: linear
 - training_steps: 10
 
-### Training results
-
-| Training Loss | Epoch | Step | Validation Loss | Cer | Wer | Mer | Wil | Wip | Hits | Substitutions | Deletions | Insertions |
-|:-------------:|:-----:|:----:|:---------------:|:---:|:---:|:---:|:---:|:---:|:----:|:-------------:|:---------:|:----------:|
-| 8.0922        | 0.0   | 4    | 7.9783          | 1.0 | 1.0 | 1.0 | 1   | 0   | 0    | 0             | 34        | 0          |
-| 7.9733        | 0.01  | 8    | 7.9280          | 1.0 | 1.0 | 1.0 | 1   | 0   | 0    | 0             | 34        | 0          |
-
-
 ### Framework versions
 
 - Transformers 4.39.3
 - Pytorch 2.1.0
 - Datasets 2.18.0
 - Tokenizers 0.15.2
-
-### Wandb run
-https://wandb.ai/alexander-polok/huggingface/runs/imoe89wk

config.json

@@ -1,18 +1,20 @@
 {
+  "_name_or_path": "Lakoc/DeCRED_small_cv",
   "architectures": [
     "JointCTCAttentionEncoderDecoder"
   ],
   "bos_token_id": 0,
-  "ctc_weight": 0.0,
+  "ctc_weight": 0.3,
   "decoder": {
-    "_name_or_path": "Lakoc/gpt2_tiny_decoder_6_layers",
+    "_name_or_path": "Lakoc/gpt2_256h_6l_add_head3_04",
     "activation_function": "gelu_new",
     "add_cross_attention": true,
     "architectures": null,
     "attn_pdrop": 0.1,
+    "average_logits": false,
     "bad_words_ids": null,
     "begin_suppress_tokens": null,
-    "bos_token_id": 50256,
+    "bos_token_id": 0,
     "chunk_size_feed_forward": 0,
     "cross_attention_hidden_size": null,
     "decoder_start_token_id": null,
@@ -21,11 +23,18 @@
     "early_stopping": false,
     "embd_pdrop": 0.1,
     "encoder_no_repeat_ngram_size": 0,
-    "eos_token_id": 50256,
+    "eos_token_id": 1,
     "exponential_decay_length_penalty": null,
     "finetuning_task": null,
     "forced_bos_token_id": null,
     "forced_eos_token_id": null,
+    "head_locations": [
+      3
+    ],
+    "head_weights": [
+      0.6,
+      0.4
+    ],
     "id2label": {
       "0": "LABEL_0",
       "1": "LABEL_1"
@@ -41,7 +50,7 @@
     "length_penalty": 1.0,
     "max_length": 20,
     "min_length": 0,
-    "model_type": "gpt2",
+    "model_type": "gpt2-multi-head",
     "n_embd": 256,
     "n_head": 4,
     "n_inner": 2048,
@@ -52,7 +61,6 @@
     "num_beams": 1,
     "num_return_sequences": 1,
     "output_attentions": false,
-    "output_hidden_size": 256,
     "output_hidden_states": false,
     "output_scores": false,
     "pad_token_id": null,
@@ -78,6 +86,7 @@
     "task_specific_params": null,
     "temperature": 1.0,
     "tf_legacy_loss": false,
+    "tie_additional_weights": false,
     "tie_encoder_decoder": false,
     "tie_word_embeddings": false,
     "tokenizer_class": null,
@@ -88,7 +97,7 @@
     "typical_p": 1.0,
     "use_bfloat16": false,
     "use_cache": true,
-    "vocab_size": 5000
+    "vocab_size": 1000
   },
   "decoder_start_token_id": 0,
   "encoder": {
@@ -251,12 +260,12 @@
     "use_fbanks": true,
     "use_macaron_ff": true,
     "use_weighted_layer_sum": false,
-    "vocab_size": 5000,
+    "vocab_size": 1000,
     "xvector_output_dim": 512
   },
   "eos_token_id": 1,
   "is_encoder_decoder": true,
-  "lsm_factor": 0,
+  "lsm_factor": 0.0,
   "mask_token_id": 4,
   "model_type": "joint_aed_ctc_speech-encoder-decoder",
   "pad_token_id": 3,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0c8461ac73ace17c62c9399d9ef6973f5cb369023b9f0c7929a8c825be0f4058
-size 155523256
+oid sha256:497190bb024e75b6d0e560b39282d01acefc76355f3940c66b2ac852a3d3a9c5
+size 144243216

predictions_test_wer94.29.csv

@@ -0,0 +1,3 @@
+label,prediction
+however due to theslow communication channels styles in the west could lag behind by to year,however
+all nouns alongside the wordsie for you always begin with a capital letter even in the middle of asentence,all

predictions_test_wer94.29_hyp.trn

@@ -0,0 +1,2 @@
+however (utterance_0)
+all (utterance_1)

predictions_test_wer94.29_hyp.trn.dtl

@@ -0,0 +1,102 @@
+DETAILED OVERALL REPORT FOR THE SYSTEM: experiment/predictions_test_wer94.29_hyp.trn
+
+SENTENCE RECOGNITION PERFORMANCE
+
+ sentences                                           2
+ with errors                            100.0%   (   2)
+
+   with substitutions                     0.0%   (   0)
+   with deletions                       100.0%   (   2)
+   with insertions                        0.0%   (   0)
+
+
+WORD RECOGNITION PERFORMANCE
+
+Percent Total Error       =   94.3%   (  33)
+
+Percent Correct           =    5.7%   (   2)
+
+Percent Substitution      =    0.0%   (   0)
+Percent Deletions         =   94.3%   (  33)
+Percent Insertions        =    0.0%   (   0)
+Percent Word Accuracy     =    5.7%
+
+
+Ref. words                =           (  35)
+Hyp. words                =           (   2)
+Aligned words             =           (  35)
+
+CONFUSION PAIRS                  Total                 (0)
+                                 With >=  1 occurrences (0)
+
+     -------
+         0
+
+
+
+INSERTIONS                       Total                 (0)
+                                 With >=  1 occurrences (0)
+
+     -------
+         0
+
+
+
+DELETIONS                        Total                 (29)
+                                 With >=  1 occurrences (29)
+
+   1:    3  ->  the
+   2:    2  ->  in
+   3:    2  ->  to
+   4:    1  ->  a
+   5:    1  ->  alongside
+   6:    1  ->  always
+   7:    1  ->  asentence
+   8:    1  ->  begin
+   9:    1  ->  behind
+  10:    1  ->  by
+  11:    1  ->  capital
+  12:    1  ->  channels
+  13:    1  ->  communication
+  14:    1  ->  could
+  15:    1  ->  due
+  16:    1  ->  even
+  17:    1  ->  for
+  18:    1  ->  lag
+  19:    1  ->  letter
+  20:    1  ->  middle
+  21:    1  ->  nouns
+  22:    1  ->  of
+  23:    1  ->  styles
+  24:    1  ->  theslow
+  25:    1  ->  west
+  26:    1  ->  with
+  27:    1  ->  wordsie
+  28:    1  ->  year
+  29:    1  ->  you
+     -------
+        33
+
+
+
+SUBSTITUTIONS                    Total                 (0)
+                                 With >=  1 occurrences (0)
+
+     -------
+         0
+
+
+* NOTE: The 'Substitution' words are those reference words
+        for which the recognizer supplied an incorrect word.
+
+
+FALSELY RECOGNIZED               Total                 (0)
+                                 With >=  1 occurrences (0)
+
+     -------
+         0
+
+
+* NOTE: The 'Falsely Recognized' words are those hypothesis words
+        which the recognizer incorrectly substituted for a reference word.
+

predictions_test_wer94.29_hyp.trn.snt.utt

@@ -0,0 +1,47 @@
+===============================================================================
+
+                     SENTENCE LEVEL REPORT FOR THE SYSTEM:
+    Name: experiment/predictions_test_wer94.29_hyp.trn
+
+===============================================================================
+
+
+SPEAKER utt
+id: (utterance_0)
+Scores: (#C #S #D #I) 1 0 15 0
+REF:  however DUE TO THESLOW COMMUNICATION CHANNELS STYLES IN THE WEST COULD LAG BEHIND BY TO YEAR 
+HYP:  however *** ** ******* ************* ******** ****** ** *** **** ***** *** ****** ** ** **** 
+Eval:         D   D  D       D             D        D      D  D   D    D     D   D      D  D  D    
+
+Correct               =    6.3%    1   (  1)
+Substitutions         =    0.0%    0   (  0)
+Deletions             =   93.8%   15   ( 15)
+Insertions            =    0.0%    0   (  0)
+
+Errors                =   93.8%   15   ( 15)
+
+Ref. words            =           16   ( 16)
+Hyp. words            =            1   (  1)
+Aligned words         =           16   ( 16)
+
+-------------------------------------------------------------------------------
+
+id: (utterance_1)
+Scores: (#C #S #D #I) 1 0 18 0
+REF:  all NOUNS ALONGSIDE THE WORDSIE FOR YOU ALWAYS BEGIN WITH A CAPITAL LETTER EVEN IN THE MIDDLE OF ASENTENCE 
+HYP:  all ***** ********* *** ******* *** *** ****** ***** **** * ******* ****** **** ** *** ****** ** ********* 
+Eval:     D     D         D   D       D   D   D      D     D    D D       D      D    D  D   D      D  D         
+
+Correct               =    5.3%    1   (  2)
+Substitutions         =    0.0%    0   (  0)
+Deletions             =   94.7%   18   ( 33)
+Insertions            =    0.0%    0   (  0)
+
+Errors                =   94.7%   18   ( 33)
+
+Ref. words            =           19   ( 35)
+Hyp. words            =            1   (  2)
+Aligned words         =           19   ( 35)
+
+-------------------------------------------------------------------------------
+

predictions_test_wer94.29_hyp.trn.sys

@@ -0,0 +1,18 @@
+
+
+
+                     SYSTEM SUMMARY PERCENTAGES by SPEAKER                      
+
+       ,----------------------------------------------------------------.
+       |          experiment/predictions_test_wer94.29_hyp.trn          |
+       |----------------------------------------------------------------|
+       | SPKR   | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err |
+       |--------+-------------+-----------------------------------------|
+       | utt    |    2     35 |  5.7    0.0   94.3    0.0   94.3  100.0 |
+       |================================================================|
+       | Sum/Avg|    2     35 |  5.7    0.0   94.3    0.0   94.3  100.0 |
+       |================================================================|
+       |  Mean  |  2.0   35.0 |  5.7    0.0   94.3    0.0   94.3  100.0 |
+       |  S.D.  |  0.0    0.0 |  0.0    0.0    0.0    0.0    0.0    0.0 |
+       | Median |  2.0   35.0 |  5.7    0.0   94.3    0.0   94.3  100.0 |
+       `----------------------------------------------------------------'

predictions_test_wer94.29_ref.trn

@@ -0,0 +1,2 @@
+however due to theslow communication channels styles in the west could lag behind by to year (utterance_0)
+all nouns alongside the wordsie for you always begin with a capital letter even in the middle of asentence (utterance_1)

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:feceb463d3f87bbaf72465b1ebe7c4dd868138ed35e59376c411884e9781847a
+oid sha256:5a87a2fc73ae493c25b7bc564a0d09f3075726207589852176ef7850651bba67
 size 5496