Initial upload

README.md

@@ -0,0 +1,57 @@
+---
+license: mit
+tags:
+- generated_from_trainer
+datasets:
+- german_dpr
+model-index:
+- name: german-qg-t5-quad
+  results: []
+---
+
+# t5-base-pretrained-qg
+
+This model is fine-tuned in question generation in German. The expected answer must be highlighted with `<hl>` token.
+
+## Task example
+
+#### Input
+```
+generate question: Obwohl die Vereinigten Staaten wie auch viele Staaten des Commonwealth Erben des <hl>britischen Common Laws<hl> sind, setzt sich das amerikanische Recht bedeutend davon ab. Dies rührt größtenteils von dem langen Zeitraum her, [...]
+```
+#### Expected output
+```
+Von welchem Gesetzt stammt das Amerikanische ab? 
+```
+
+## Model description
+
+This model is a fine-tuned version of [valhalla/t5-base-qg-hl](https://huggingface.co/valhalla/t5-base-qg-hl) on the [GermanQUAD](https://www.deepset.ai/germanquad) dataset.
+
+## Training and evaluation data
+
+The training script can be accessed [here](https://github.com/d-e-h-i-o/german-qg).
+
+### Evaluation
+
+The model achieves a BLEU-4 score of **11.30** on the GermanQuAD test set (n=2204). 
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 2
+- eval_batch_size: 2
+- seed: 100
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- num_epochs: 10
+
+### Framework versions
+
+- Transformers 4.13.0.dev0
+- Pytorch 1.10.0+cu102
+- Datasets 1.16.1
+- Tokenizers 0.10.3

added_tokens.json

@@ -0,0 +1 @@
+{"<hl>": 32100}

all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 10.0,
+    "predict_bleu": 13.4716,
+    "predict_gen_len": 16.7945,
+    "predict_loss": 1.5842393636703491,
+    "predict_runtime": 384.6342,
+    "predict_samples": 2204,
+    "predict_samples_per_second": 5.73,
+    "predict_steps_per_second": 2.865,
+    "train_loss": 0.9689414988638463,
+    "train_runtime": 6282.5026,
+    "train_samples": 11518,
+    "train_samples_per_second": 18.333,
+    "train_steps_per_second": 1.144
+}

config.json

@@ -0,0 +1,54 @@
+{
+  "_name_or_path": "dehio/german-qg-t5-quad",
+  "architectures": [
+    "T5ForConditionalGeneration"
+  ],
+  "d_ff": 3072,
+  "d_kv": 64,
+  "d_model": 768,
+  "decoder_start_token_id": 0,
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "relu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "n_positions": 512,
+  "num_decoder_layers": 12,
+  "num_heads": 12,
+  "num_layers": 12,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_num_buckets": 32,
+  "task_specific_params": {
+    "summarization": {
+      "early_stopping": true,
+      "max_length": 32,
+      "num_beams": 4,
+      "prefix": ""
+    },
+    "translation_en_to_de": {
+      "early_stopping": true,
+      "max_length": 32,
+      "num_beams": 4,
+      "prefix": ""
+    },
+    "translation_en_to_fr": {
+      "early_stopping": true,
+      "max_length": 32,
+      "num_beams": 4,
+      "prefix": ""
+    },
+    "translation_en_to_ro": {
+      "early_stopping": true,
+      "max_length": 32,
+      "num_beams": 4,
+      "prefix": ""
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.13.0.dev0",
+  "use_cache": true,
+  "vocab_size": 32101
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79b661c5955b1c3ecaada3bb43f35502b8107bf7313e97a177006f6244d14fc6
+size 891647935

special_tokens_map.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "additional_special_tokens": ["<extra_id_0>", "<extra_id_1>", "<extra_id_2>", "<extra_id_3>", "<extra_id_4>", "<extra_id_5>", "<extra_id_6>", "<extra_id_7>", "<extra_id_8>", "<extra_id_9>", "<extra_id_10>", "<extra_id_11>", "<extra_id_12>", "<extra_id_13>", "<extra_id_14>", "<extra_id_15>", "<extra_id_16>", "<extra_id_17>", "<extra_id_18>", "<extra_id_19>", "<extra_id_20>", "<extra_id_21>", "<extra_id_22>", "<extra_id_23>", "<extra_id_24>", "<extra_id_25>", "<extra_id_26>", "<extra_id_27>", "<extra_id_28>", "<extra_id_29>", "<extra_id_30>", "<extra_id_31>", "<extra_id_32>", "<extra_id_33>", "<extra_id_34>", "<extra_id_35>", "<extra_id_36>", "<extra_id_37>", "<extra_id_38>", "<extra_id_39>", "<extra_id_40>", "<extra_id_41>", "<extra_id_42>", "<extra_id_43>", "<extra_id_44>", "<extra_id_45>", "<extra_id_46>", "<extra_id_47>", "<extra_id_48>", "<extra_id_49>", "<extra_id_50>", "<extra_id_51>", "<extra_id_52>", "<extra_id_53>", "<extra_id_54>", "<extra_id_55>", "<extra_id_56>", "<extra_id_57>", "<extra_id_58>", "<extra_id_59>", "<extra_id_60>", "<extra_id_61>", "<extra_id_62>", "<extra_id_63>", "<extra_id_64>", "<extra_id_65>", "<extra_id_66>", "<extra_id_67>", "<extra_id_68>", "<extra_id_69>", "<extra_id_70>", "<extra_id_71>", "<extra_id_72>", "<extra_id_73>", "<extra_id_74>", "<extra_id_75>", "<extra_id_76>", "<extra_id_77>", "<extra_id_78>", "<extra_id_79>", "<extra_id_80>", "<extra_id_81>", "<extra_id_82>", "<extra_id_83>", "<extra_id_84>", "<extra_id_85>", "<extra_id_86>", "<extra_id_87>", "<extra_id_88>", "<extra_id_89>", "<extra_id_90>", "<extra_id_91>", "<extra_id_92>", "<extra_id_93>", "<extra_id_94>", "<extra_id_95>", "<extra_id_96>", "<extra_id_97>", "<extra_id_98>", "<extra_id_99>"]}

spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d60acb128cf7b7f2536e8f38a5b18a05535c9e14c7a355904270e15b0945ea86
+size 791656

test_results.json

@@ -0,0 +1,9 @@
+{
+    "predict_bleu": 13.4716,
+    "predict_gen_len": 16.7945,
+    "predict_loss": 1.5842393636703491,
+    "predict_runtime": 384.6342,
+    "predict_samples": 2204,
+    "predict_samples_per_second": 5.73,
+    "predict_steps_per_second": 2.865
+}

tokenizer_config.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "extra_ids": 100, "additional_special_tokens": ["<extra_id_0>", "<extra_id_1>", "<extra_id_2>", "<extra_id_3>", "<extra_id_4>", "<extra_id_5>", "<extra_id_6>", "<extra_id_7>", "<extra_id_8>", "<extra_id_9>", "<extra_id_10>", "<extra_id_11>", "<extra_id_12>", "<extra_id_13>", "<extra_id_14>", "<extra_id_15>", "<extra_id_16>", "<extra_id_17>", "<extra_id_18>", "<extra_id_19>", "<extra_id_20>", "<extra_id_21>", "<extra_id_22>", "<extra_id_23>", "<extra_id_24>", "<extra_id_25>", "<extra_id_26>", "<extra_id_27>", "<extra_id_28>", "<extra_id_29>", "<extra_id_30>", "<extra_id_31>", "<extra_id_32>", "<extra_id_33>", "<extra_id_34>", "<extra_id_35>", "<extra_id_36>", "<extra_id_37>", "<extra_id_38>", "<extra_id_39>", "<extra_id_40>", "<extra_id_41>", "<extra_id_42>", "<extra_id_43>", "<extra_id_44>", "<extra_id_45>", "<extra_id_46>", "<extra_id_47>", "<extra_id_48>", "<extra_id_49>", "<extra_id_50>", "<extra_id_51>", "<extra_id_52>", "<extra_id_53>", "<extra_id_54>", "<extra_id_55>", "<extra_id_56>", "<extra_id_57>", "<extra_id_58>", "<extra_id_59>", "<extra_id_60>", "<extra_id_61>", "<extra_id_62>", "<extra_id_63>", "<extra_id_64>", "<extra_id_65>", "<extra_id_66>", "<extra_id_67>", "<extra_id_68>", "<extra_id_69>", "<extra_id_70>", "<extra_id_71>", "<extra_id_72>", "<extra_id_73>", "<extra_id_74>", "<extra_id_75>", "<extra_id_76>", "<extra_id_77>", "<extra_id_78>", "<extra_id_79>", "<extra_id_80>", "<extra_id_81>", "<extra_id_82>", "<extra_id_83>", "<extra_id_84>", "<extra_id_85>", "<extra_id_86>", "<extra_id_87>", "<extra_id_88>", "<extra_id_89>", "<extra_id_90>", "<extra_id_91>", "<extra_id_92>", "<extra_id_93>", "<extra_id_94>", "<extra_id_95>", "<extra_id_96>", "<extra_id_97>", "<extra_id_98>", "<extra_id_99>"], "model_max_length": 512, "special_tokens_map_file": "models/german-qg-t5/special_tokens_map.json", "full_tokenizer_file": null, "name_or_path": "dehio/german-qg-t5-quad", "sp_model_kwargs": {}, "tokenizer_class": "T5Tokenizer"}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 10.0,
+    "train_loss": 0.9689414988638463,
+    "train_runtime": 6282.5026,
+    "train_samples": 11518,
+    "train_samples_per_second": 18.333,
+    "train_steps_per_second": 1.144
+}

trainer_state.json

@@ -0,0 +1,451 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.998784511199862,
+  "global_step": 7190,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.860917941585536e-05,
+      "loss": 1.7018,
+      "step": 100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.721835883171071e-05,
+      "loss": 1.5678,
+      "step": 200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.582753824756607e-05,
+      "loss": 1.5333,
+      "step": 300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.443671766342142e-05,
+      "loss": 1.5235,
+      "step": 400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.304589707927678e-05,
+      "loss": 1.4892,
+      "step": 500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.165507649513213e-05,
+      "loss": 1.4781,
+      "step": 600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.026425591098748e-05,
+      "loss": 1.4319,
+      "step": 700
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.887343532684285e-05,
+      "loss": 1.3011,
+      "step": 800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.74826147426982e-05,
+      "loss": 1.2885,
+      "step": 900
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.609179415855355e-05,
+      "loss": 1.2524,
+      "step": 1000
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.47009735744089e-05,
+      "loss": 1.2674,
+      "step": 1100
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.331015299026426e-05,
+      "loss": 1.2533,
+      "step": 1200
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.191933240611962e-05,
+      "loss": 1.268,
+      "step": 1300
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.052851182197498e-05,
+      "loss": 1.2499,
+      "step": 1400
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.913769123783032e-05,
+      "loss": 1.1782,
+      "step": 1500
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.774687065368567e-05,
+      "loss": 1.1113,
+      "step": 1600
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.635605006954103e-05,
+      "loss": 1.1167,
+      "step": 1700
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.496522948539638e-05,
+      "loss": 1.1422,
+      "step": 1800
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.357440890125175e-05,
+      "loss": 1.118,
+      "step": 1900
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.21835883171071e-05,
+      "loss": 1.1133,
+      "step": 2000
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.079276773296244e-05,
+      "loss": 1.1058,
+      "step": 2100
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.94019471488178e-05,
+      "loss": 1.0747,
+      "step": 2200
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.801112656467315e-05,
+      "loss": 1.0059,
+      "step": 2300
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.662030598052852e-05,
+      "loss": 0.9988,
+      "step": 2400
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.522948539638388e-05,
+      "loss": 1.0111,
+      "step": 2500
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.383866481223923e-05,
+      "loss": 1.0237,
+      "step": 2600
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.244784422809457e-05,
+      "loss": 1.0011,
+      "step": 2700
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.105702364394992e-05,
+      "loss": 1.0133,
+      "step": 2800
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.966620305980529e-05,
+      "loss": 0.9724,
+      "step": 2900
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.827538247566065e-05,
+      "loss": 0.9005,
+      "step": 3000
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.6884561891515995e-05,
+      "loss": 0.9264,
+      "step": 3100
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.549374130737135e-05,
+      "loss": 0.933,
+      "step": 3200
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.41029207232267e-05,
+      "loss": 0.9283,
+      "step": 3300
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.2712100139082064e-05,
+      "loss": 0.9033,
+      "step": 3400
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.132127955493742e-05,
+      "loss": 0.9139,
+      "step": 3500
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.993045897079277e-05,
+      "loss": 0.9383,
+      "step": 3600
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.853963838664812e-05,
+      "loss": 0.8268,
+      "step": 3700
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.714881780250348e-05,
+      "loss": 0.841,
+      "step": 3800
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.5757997218358836e-05,
+      "loss": 0.8594,
+      "step": 3900
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.436717663421418e-05,
+      "loss": 0.8346,
+      "step": 4000
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.2976356050069544e-05,
+      "loss": 0.8626,
+      "step": 4100
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.15855354659249e-05,
+      "loss": 0.8691,
+      "step": 4200
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.019471488178025e-05,
+      "loss": 0.8424,
+      "step": 4300
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.880389429763561e-05,
+      "loss": 0.793,
+      "step": 4400
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.741307371349096e-05,
+      "loss": 0.8025,
+      "step": 4500
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6022253129346316e-05,
+      "loss": 0.8245,
+      "step": 4600
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.463143254520167e-05,
+      "loss": 0.797,
+      "step": 4700
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3240611961057024e-05,
+      "loss": 0.8063,
+      "step": 4800
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.184979137691238e-05,
+      "loss": 0.8042,
+      "step": 4900
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0458970792767733e-05,
+      "loss": 0.8068,
+      "step": 5000
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.906815020862309e-05,
+      "loss": 0.7873,
+      "step": 5100
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.767732962447844e-05,
+      "loss": 0.7495,
+      "step": 5200
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6286509040333796e-05,
+      "loss": 0.7538,
+      "step": 5300
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4895688456189153e-05,
+      "loss": 0.7631,
+      "step": 5400
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3504867872044508e-05,
+      "loss": 0.7532,
+      "step": 5500
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2114047287899862e-05,
+      "loss": 0.7772,
+      "step": 5600
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0723226703755216e-05,
+      "loss": 0.7691,
+      "step": 5700
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.933240611961057e-05,
+      "loss": 0.7544,
+      "step": 5800
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7941585535465928e-05,
+      "loss": 0.7243,
+      "step": 5900
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.655076495132128e-05,
+      "loss": 0.7308,
+      "step": 6000
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.5159944367176635e-05,
+      "loss": 0.7115,
+      "step": 6100
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.376912378303199e-05,
+      "loss": 0.7275,
+      "step": 6200
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2378303198887344e-05,
+      "loss": 0.7279,
+      "step": 6300
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.0987482614742698e-05,
+      "loss": 0.7497,
+      "step": 6400
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.596662030598054e-06,
+      "loss": 0.7163,
+      "step": 6500
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.205841446453408e-06,
+      "loss": 0.7047,
+      "step": 6600
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.815020862308763e-06,
+      "loss": 0.7089,
+      "step": 6700
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.424200278164117e-06,
+      "loss": 0.7078,
+      "step": 6800
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.033379694019471e-06,
+      "loss": 0.7065,
+      "step": 6900
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.6425591098748263e-06,
+      "loss": 0.7017,
+      "step": 7000
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.2517385257301808e-06,
+      "loss": 0.6855,
+      "step": 7100
+    },
+    {
+      "epoch": 10.0,
+      "step": 7190,
+      "total_flos": 6.20936877252096e+16,
+      "train_loss": 0.9689414988638463,
+      "train_runtime": 6282.5026,
+      "train_samples_per_second": 18.333,
+      "train_steps_per_second": 1.144
+    }
+  ],
+  "max_steps": 7190,
+  "num_train_epochs": 10,
+  "total_flos": 6.20936877252096e+16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc3e1f465cde2661ed9562251787718f1ce3e2112920cf5a88b7c91a5d00224e
+size 3055