Delete training_params.json as the training info on the Readme explains it better
Browse files- training_params.json +0 -47
training_params.json
DELETED
|
@@ -1,47 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"model": "./TinyMistral-248M/",
|
| 3 |
-
"data_path": "OpenAssistant/oasst_top1_2023-08-25",
|
| 4 |
-
"project_name": "trained-model",
|
| 5 |
-
"train_split": "train",
|
| 6 |
-
"valid_split": "test",
|
| 7 |
-
"text_column": "text",
|
| 8 |
-
"rejected_text_column": "rejected",
|
| 9 |
-
"token": null,
|
| 10 |
-
"lr": 1e-05,
|
| 11 |
-
"epochs": 5,
|
| 12 |
-
"batch_size": 2,
|
| 13 |
-
"warmup_ratio": 0.05,
|
| 14 |
-
"gradient_accumulation": 8,
|
| 15 |
-
"optimizer": "adamw_torch",
|
| 16 |
-
"scheduler": "constant",
|
| 17 |
-
"weight_decay": 0.0,
|
| 18 |
-
"max_grad_norm": 1.0,
|
| 19 |
-
"seed": 42,
|
| 20 |
-
"add_eos_token": false,
|
| 21 |
-
"block_size": 1024,
|
| 22 |
-
"use_peft": false,
|
| 23 |
-
"lora_r": 16,
|
| 24 |
-
"lora_alpha": 32,
|
| 25 |
-
"lora_dropout": 0.05,
|
| 26 |
-
"logging_steps": 50,
|
| 27 |
-
"evaluation_strategy": "steps",
|
| 28 |
-
"save_total_limit": 2,
|
| 29 |
-
"save_strategy": "steps",
|
| 30 |
-
"auto_find_batch_size": false,
|
| 31 |
-
"fp16": false,
|
| 32 |
-
"push_to_hub": false,
|
| 33 |
-
"use_int8": false,
|
| 34 |
-
"model_max_length": 4096,
|
| 35 |
-
"repo_id": null,
|
| 36 |
-
"use_int4": false,
|
| 37 |
-
"trainer": "sft",
|
| 38 |
-
"target_modules": null,
|
| 39 |
-
"merge_adapter": false,
|
| 40 |
-
"username": null,
|
| 41 |
-
"use_flash_attention_2": false,
|
| 42 |
-
"log": "none",
|
| 43 |
-
"disable_gradient_checkpointing": false,
|
| 44 |
-
"model_ref": null,
|
| 45 |
-
"dpo_beta": 0.1,
|
| 46 |
-
"prompt_text_column": "prompt"
|
| 47 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|