Hugging Face's logo

ibm-granite
/
granite-3.2-8b-lora-rag-hallucination-detection

PEFT
Safetensors
Model card Files Community
1
granite-3.2-8b-lora-rag-hallucination-detection
File size: 2,697 Bytes 
c5f698f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
 
{
    "rl_model_name": null,
    "rl_model_class": null,
    "num_rollouts": 64,
    "chunk_size": 16,
    "ppo_epochs": 4,
    "init_kl_coef": 0.05,
    "num_layers_unfrozen": 2,
    "rm_bits": 8,
    "rl_lora_rank": 8,
    "rl_lora_alpha": 32,
    "rl_lora_dropout": 0.1,
    "use_qlora_in_rl": false,
    "use_rl_peft_config": true,
    "lora_compute_dtype": "bfloat16",
    "steps_per_print": 10,
    "logdir": "aim-repo",
    "aim_repo": null,
    "experiment_name": "granite_3.2_faithfulness_balanced_response_id",
    "stage": 2,
    "overlap_comm": false,
    "contiguous_gradients": false,
    "cpu_offload": false,
    "optimizer": {
        "optimizer_class": "FusedAdam",
        "lr": 1e-05,
        "weight_decay": 0.1,
        "betas": [
            0.9,
            0.95
        ],
        "eps": 1e-10
    },
    "lr_schedule": "linear",
    "warmup_steps": 200,
    "datasets": [
        {
            "data_class": "JSONLinesDatasetRAGChat",
            "data_name": "simulator-citations-chat",
            "data_path": "data/faithfulness/full_response_balanced_response_id",
            "data_sampling_proportion": 1,
            "max_input_tokens": 5000,
            "max_output_tokens": 500
        }
    ],
    "seed": 42,
    "training_inference_type": "lora_finetuning",
    "prompt_tuning_init": null,
    "prompt_tuning_init_text": null,
    "num_virtual_tokens": null,
    "load_path": null,
    "peft_num_layers": null,
    "peft_num_attention_heads": null,
    "model_name": "ibm-granite/granite-3.2-8b-instruct",
    "tokenizer_name": null,
    "model_class": "AutoModelForCausalLM",
    "gma_model_class": "Model",
    "dtype": "bfloat16",
    "trust_remote_code": false,
    "padding_side": null,
    "lora_rank": 8,
    "lora_alpha": 8,
    "lora_dropout": 0.1,
    "lora_target_modules": [
        "q_proj",
        "k_proj",
        "v_proj",
        "o_proj",
        "up_proj",
        "down_proj",
        "gate_proj"
    ],
    "save_huggingface_checkpoint": true,
    "quantization_method": "fp4",
    "bnb_4bit_use_double_quant": false,
    "use_quantization_for_inference": false,
    "max_seq_len": 2048,
    "attention_implementation": "flash_attention_2",
    "num_labels": 1,
    "use_sdpa_attention": false,
    "save_path": "checkpoints/granite_3.2_faithfulness_balanced_response_id",
    "ignore_sampling_proportion_for_validation": false,
    "num_training_steps": 200000,
    "gradient_accumulation_steps": 1,
    "eval_interval": 20000,
    "save_interval": 20000,
    "batch_size_per_gpu": 1,
    "coeff": 1.0,
    "eval_during_training": true,
    "smart_token_allocation": false,
    "max_new_tokens": 0,
    "gradient_checkpointing": true
}