rulm

1
{
2
    "trainer": {
3
        "evaluation_strategy": "steps",
4
        "per_device_train_batch_size": 1,
5
        "per_device_eval_batch_size": 2,
6
        "gradient_accumulation_steps": 32,
7
        "eval_steps": 150,
8
        "save_steps": 150,
9
        "logging_steps": 10,
10
        "learning_rate": 0.0003,
11
        "num_train_epochs": 3,
12
        "lr_scheduler_type": "cosine",
13
        "warmup_steps": 100,
14
        "fp16": false,
15
        "bf16": false,
16
        "torch_compile": false,
17
        "optim": "adamw_torch"
18
    },
19
    "lora": {
20
        "r": 4,
21
        "lora_alpha": 32,
22
        "lora_dropout": 0.02,
23
        "inference_mode": false,
24
        "task_type": "SEQ_2_SEQ_LM"
25
    },
26
    "load_in_8bit": true,
27
    "model_name": "bigscience/mt0-xxl-mt",
28
    "model_type": "seq2seq",
29
    "template_category": "seq2seq_no_newlines",
30
    "max_source_tokens_count": 400,
31
    "max_target_tokens_count": 400
32
}
33

34