Upload training_parameters.json
Browse files- training_parameters.json +37 -0
training_parameters.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"lora_name": "TinyTR",
|
| 3 |
+
"always_override": false,
|
| 4 |
+
"save_steps": 350.0,
|
| 5 |
+
"micro_batch_size": 6,
|
| 6 |
+
"batch_size": 0,
|
| 7 |
+
"epochs": 0.25,
|
| 8 |
+
"learning_rate": "3e-4",
|
| 9 |
+
"lr_scheduler_type": "linear",
|
| 10 |
+
"lora_rank": 64,
|
| 11 |
+
"lora_alpha": 128,
|
| 12 |
+
"lora_dropout": 0.1,
|
| 13 |
+
"cutoff_len": 768,
|
| 14 |
+
"dataset": "turkish",
|
| 15 |
+
"eval_dataset": "None",
|
| 16 |
+
"format": "Chatml",
|
| 17 |
+
"eval_steps": 0.0,
|
| 18 |
+
"raw_text_file": "None",
|
| 19 |
+
"higher_rank_limit": false,
|
| 20 |
+
"warmup_steps": 120.0,
|
| 21 |
+
"optimizer": "adamw_torch",
|
| 22 |
+
"hard_cut_string": "\\n\\n\\n",
|
| 23 |
+
"train_only_after": "",
|
| 24 |
+
"stop_at_loss": 0,
|
| 25 |
+
"add_eos_token": true,
|
| 26 |
+
"min_chars": 0.0,
|
| 27 |
+
"report_to": "None",
|
| 28 |
+
"precize_slicing_overlap": true,
|
| 29 |
+
"add_eos_token_type": "Every Block",
|
| 30 |
+
"save_steps_under_loss": 0,
|
| 31 |
+
"add_bos_token": true,
|
| 32 |
+
"training_projection": "q-k-v-o",
|
| 33 |
+
"sliding_window": false,
|
| 34 |
+
"warmup_ratio": 0,
|
| 35 |
+
"grad_accumulation": 2,
|
| 36 |
+
"neft_noise_alpha": 6
|
| 37 |
+
}
|