Pieter Delobelle
commited on
Commit
•
b66b2e7
1
Parent(s):
cfd0bb3
Initial commit
Browse files- config.json +23 -0
- merges.txt +0 -0
- parameters.json +51 -0
- pytorch_model.bin +3 -0
- vocab.json +0 -0
config.json
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"attention_probs_dropout_prob": 0.1,
|
| 3 |
+
"bos_token_id": 0,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"gradient_checkpointing": false,
|
| 6 |
+
"hidden_act": "gelu",
|
| 7 |
+
"hidden_dropout_prob": 0.1,
|
| 8 |
+
"hidden_size": 768,
|
| 9 |
+
"initializer_range": 0.02,
|
| 10 |
+
"intermediate_size": 3072,
|
| 11 |
+
"layer_norm_eps": 1e-05,
|
| 12 |
+
"max_position_embeddings": 514,
|
| 13 |
+
"model_type": "roberta",
|
| 14 |
+
"num_attention_heads": 12,
|
| 15 |
+
"num_hidden_layers": 6,
|
| 16 |
+
"output_hidden_states": true,
|
| 17 |
+
"pad_token_id": 1,
|
| 18 |
+
"position_embedding_type": "absolute",
|
| 19 |
+
"transformers_version": "4.3.3",
|
| 20 |
+
"type_vocab_size": 1,
|
| 21 |
+
"use_cache": true,
|
| 22 |
+
"vocab_size": 40000
|
| 23 |
+
}
|
merges.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
parameters.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"force": false,
|
| 3 |
+
"dump_path": "serialization_dir/distilrobbert-v2-mini-unshuffled-merged-0.5",
|
| 4 |
+
"data_file": "data/oscar_dutch/nl_unshuffled_mini_0.5-data.pdelobelle-robbert-v2-dutch-base.pickle",
|
| 5 |
+
"student_type": "roberta",
|
| 6 |
+
"student_config": "training_configs/distilrobbert-base.json",
|
| 7 |
+
"student_pretrained_weights": null,
|
| 8 |
+
"teacher_type": "roberta",
|
| 9 |
+
"teacher_name": "pdelobelle/robbert-v2-dutch-base",
|
| 10 |
+
"temperature": 2.0,
|
| 11 |
+
"alpha_ce": 5.0,
|
| 12 |
+
"alpha_mlm": 2.0,
|
| 13 |
+
"alpha_clm": 0.0,
|
| 14 |
+
"alpha_mse": 0.0,
|
| 15 |
+
"alpha_cos": 1.0,
|
| 16 |
+
"mlm": true,
|
| 17 |
+
"mlm_mask_prop": 0.15,
|
| 18 |
+
"word_mask": 0.8,
|
| 19 |
+
"word_keep": 0.1,
|
| 20 |
+
"word_rand": 0.1,
|
| 21 |
+
"mlm_smoothing": 0.7,
|
| 22 |
+
"token_counts": "data/oscar_dutch/nl_unshuffled_mini_0.5-counts.pdelobelle-robbert-v2-dutch-base.pickle",
|
| 23 |
+
"restrict_ce_to_mask": false,
|
| 24 |
+
"freeze_pos_embs": true,
|
| 25 |
+
"freeze_token_type_embds": false,
|
| 26 |
+
"n_epoch": 3,
|
| 27 |
+
"batch_size": 5,
|
| 28 |
+
"group_by_size": true,
|
| 29 |
+
"gradient_accumulation_steps": 128,
|
| 30 |
+
"warmup_prop": 0.05,
|
| 31 |
+
"weight_decay": 0.01,
|
| 32 |
+
"learning_rate": 0.0005,
|
| 33 |
+
"adam_epsilon": 1e-06,
|
| 34 |
+
"max_grad_norm": 5.0,
|
| 35 |
+
"initializer_range": 0.02,
|
| 36 |
+
"fp16": false,
|
| 37 |
+
"fp16_opt_level": "O1",
|
| 38 |
+
"gpus": 1,
|
| 39 |
+
"local_rank": 0,
|
| 40 |
+
"seed": 56,
|
| 41 |
+
"log_interval": 500,
|
| 42 |
+
"checkpoint_interval": 4000,
|
| 43 |
+
"n_nodes": 1,
|
| 44 |
+
"node_id": 0,
|
| 45 |
+
"global_rank": 0,
|
| 46 |
+
"world_size": 1,
|
| 47 |
+
"n_gpu_per_node": 1,
|
| 48 |
+
"multi_gpu": false,
|
| 49 |
+
"is_master": true,
|
| 50 |
+
"multi_node": false
|
| 51 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:78ed86e98375e44dcae6dd2a8403e8479e273124854c68af4c3f6ab094eb2ccb
|
| 3 |
+
size 297155506
|
vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|