Upload 3 files
Browse filesRename English adapter from OneM-enko-en-mean_eng to OneM-enen-en-mean_eng
OneM-enen-en-mean_eng-32.1e-4.1000/.DS_Store
ADDED
|
Binary file (6.15 kB). View file
|
|
|
OneM-enen-en-mean_eng-32.1e-4.1000/1M.enen.en_mean_eng.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2345fa23d78590ae81c3d9f141502cef02149a2fadd11bb3dc5b6bab189908dc
|
| 3 |
+
size 100923736
|
OneM-enen-en-mean_eng-32.1e-4.1000/config.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epochs": 50,
|
| 3 |
+
"batch_size": 8,
|
| 4 |
+
"valid_batch_size": 16,
|
| 5 |
+
"gradient_accumulation_steps": 1,
|
| 6 |
+
"learning_rate": 0.0001,
|
| 7 |
+
"warmup_steps": 1000,
|
| 8 |
+
"early_stop": 3,
|
| 9 |
+
"mask_rate": 0.2,
|
| 10 |
+
"seed": 2023,
|
| 11 |
+
"training_steps": 100000,
|
| 12 |
+
"eval_steps": 5000,
|
| 13 |
+
"do_adapter_train": true,
|
| 14 |
+
"do_task_finetune": false,
|
| 15 |
+
"do_predict": false,
|
| 16 |
+
"do_synthesize": false,
|
| 17 |
+
"output_dir": "output",
|
| 18 |
+
"exp_tag": "OneM-enko-en-mean_eng-32.1e-4.1000",
|
| 19 |
+
"use_checkpoint": false,
|
| 20 |
+
"adapter_training_checkpoint": null,
|
| 21 |
+
"pretrained_adapter_dir": null,
|
| 22 |
+
"model_name_or_path": "google/mt5-large",
|
| 23 |
+
"max_seq_length": 512,
|
| 24 |
+
"local_rank": -1,
|
| 25 |
+
"adapter_types": "decoder-lang",
|
| 26 |
+
"langs": "en,ko",
|
| 27 |
+
"partial_langs": "en",
|
| 28 |
+
"tasks": null,
|
| 29 |
+
"task_lang": null,
|
| 30 |
+
"adapter_layer_norm": true,
|
| 31 |
+
"adapter_reduction_factor": 2,
|
| 32 |
+
"adapter_hidden_act": "gelu",
|
| 33 |
+
"frozen_list": "all",
|
| 34 |
+
"freeze_option": "dec_cross_attention_only",
|
| 35 |
+
"train_spider_seq2seq_prefix": "data/preprocessed_data/train_spider_seq2seq",
|
| 36 |
+
"dev_spider_seq2seq_prefix": "data/preprocessed_data/dev_spider_seq2seq",
|
| 37 |
+
"without_schema": false,
|
| 38 |
+
"inference_data_file": "/home1/deokhk_1/research/ZX-seq2seq/data/preprocessed_data/dev_spider_seq2seq_english.json",
|
| 39 |
+
"generation_max_length": 64,
|
| 40 |
+
"generation_num_beams": 8,
|
| 41 |
+
"contrastive_generation": false,
|
| 42 |
+
"repetition_penalty": 1.0,
|
| 43 |
+
"without_language_identity": true,
|
| 44 |
+
"removal_type": "mean_eng",
|
| 45 |
+
"language_identity_path": "/home1/deokhk_1/research/XLang-NL2SQL/language_identity_wiki_token_1M",
|
| 46 |
+
"remove_prob": 1.0,
|
| 47 |
+
"induced_code_switch": false,
|
| 48 |
+
"gt_code_switch": false,
|
| 49 |
+
"switch_prob": 1.0,
|
| 50 |
+
"muse_dict_path": "/home1/deokhk_1/research/ZX-seq2seq/data/muse_dict.txt",
|
| 51 |
+
"induced_dict_path": "/home/deokhk/research/XLang-NL2SQL/bilingual_dictionary/ko_to_en_token_dict_csls.json",
|
| 52 |
+
"result_prefix": ""
|
| 53 |
+
}
|