Commit ·
22103f1
1
Parent(s): 65b1e66
Upload 2 files
Browse files- training_log.json +10 -0
- training_parameters.json +26 -0
training_log.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train_runtime": 665.8011,
|
| 3 |
+
"train_samples_per_second": 1.494,
|
| 4 |
+
"train_steps_per_second": 0.011,
|
| 5 |
+
"total_flos": 1.772375330783232e+16,
|
| 6 |
+
"train_loss": 2.5272955894470215,
|
| 7 |
+
"epoch": 0.9,
|
| 8 |
+
"loss": 2.6658,
|
| 9 |
+
"learning_rate": 0.00021428571428571427
|
| 10 |
+
}
|
training_parameters.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"lora_name": "jql_llama_13b",
|
| 3 |
+
"always_override": true,
|
| 4 |
+
"save_steps": 0.0,
|
| 5 |
+
"micro_batch_size": 8,
|
| 6 |
+
"batch_size": 128,
|
| 7 |
+
"epochs": 1.0,
|
| 8 |
+
"learning_rate": "3e-4",
|
| 9 |
+
"lr_scheduler_type": "linear",
|
| 10 |
+
"lora_rank": 32,
|
| 11 |
+
"lora_alpha": 64,
|
| 12 |
+
"lora_dropout": 0.05,
|
| 13 |
+
"cutoff_len": 256,
|
| 14 |
+
"dataset": "jql_chat_gpt_v2",
|
| 15 |
+
"eval_dataset": "None",
|
| 16 |
+
"format": "jql-format",
|
| 17 |
+
"eval_steps": 100.0,
|
| 18 |
+
"raw_text_file": "None",
|
| 19 |
+
"overlap_len": 128,
|
| 20 |
+
"newline_favor_len": 128,
|
| 21 |
+
"higher_rank_limit": false,
|
| 22 |
+
"warmup_steps": 100.0,
|
| 23 |
+
"optimizer": "adamw_torch",
|
| 24 |
+
"hard_cut_string": "\\n\\n\\n",
|
| 25 |
+
"train_only_after": ""
|
| 26 |
+
}
|