local --cuda out of memory == 3869MiB / 3910MiB

[RohitDhankar]

• local GPU out of memory even with -- "batch_size": 2,

{
    "model": "abhishek/llama-2-7b-hf-small-shards",
    "data_path": ".",
    "project_name": "llm_1a",
    "train_split": "train",
    "valid_split": null,
    "text_column": "text",
    "token": null,
    "lr": 0.0002,
    "epochs": 3,
    "batch_size": 2,
    "warmup_ratio": 0.1,
    "gradient_accumulation": 1,
    "optimizer": "adamw_torch",
    "scheduler": "linear",
    "weight_decay": 0.0,
    "max_grad_norm": 1.0,
    "seed": 42,
    "add_eos_token": false,
    "block_size": -1,
    "use_peft": true,
    "lora_r": 16,
    "lora_alpha": 32,
    "lora_dropout": 0.05,
    "logging_steps": -1,
    "evaluation_strategy": "epoch",
    "save_total_limit": 1,
    "save_strategy": "epoch",
    "auto_find_batch_size": false,
    "fp16": false,
    "push_to_hub": false,
    "use_int8": false,
    "model_max_length": 1024,
    "repo_id": null,
    "use_int4": true,
    "trainer": "sft",
    "target_modules": null,
    "merge_adapter": false,
    "username": null,
    "use_flash_attention_2": false
}