File size: 4,033 Bytes
a57357b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
adb15f9
 
20852a7
a57357b
 
 
 
 
 
 
 
 
 
3da7418
a57357b
 
9f8478c
 
 
a57357b
 
 
 
 
 
 
 
 
 
 
 
 
 
93b2fec
a57357b
 
 
 
 
 
 
 
 
 
 
3da7418
 
4a1fd53
3da7418
 
 
 
 
 
bf7bd7e
 
 
 
 
 
 
 
3da7418
 
a57357b
 
 
 
 
 
 
5a7635c
a57357b
 
 
 
 
 
 
9f8478c
4a1fd53
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bf7bd7e
4a1fd53
 
bf7bd7e
 
4a1fd53
 
 
 
 
 
 
 
 
 
 
 
 
 
5a7635c
5b7edf3
4a1fd53
 
 
dc055e5
 
 
 
4a1fd53
 
 
dc055e5
4a1fd53
 
 
 
 
ae57ea2
4a1fd53
 
 
 
 
 
 
 
ae57ea2
5a7635c
4a1fd53
 
5a7635c
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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
{
  "model": {
    "name": "unsloth/phi-4-unsloth-bnb-4bit",
    "trust_remote_code": true,
    "use_fast_tokenizer": true
  },

  "tokenizer": {
    "chat_template": "phi",
    "max_seq_length": 2048,
    "padding_side": "right",
    "add_eos_token": true
  },

  "training": {
    "per_device_train_batch_size": 16,
    "gradient_accumulation_steps": 3,
    "learning_rate": 2e-5,
    "num_train_epochs": 3,
    "max_steps": -1,
    "logging_steps": 10,
    "save_steps": 200,
    "save_total_limit": 5,
    "push_to_hub": true,
    "hub_strategy": "every_save",
    "gradient_checkpointing": true,
    "optim": "adamw_torch",
    "lr_scheduler_type": "cosine",
    "warmup_ratio": 0.05,
    "weight_decay": 0.01,
    "max_grad_norm": 1.0,
    "neftune_noise_alpha": 5,
    "fp16": false,
    "bf16": true
  },

  "checkpointing": {
    "output_dir": "./results",
    "save_strategy": "steps",
    "save_steps": 100,
    "save_total_limit": 3,
    "hub_strategy": "every_save"
  },

  "unsloth": {
    "enabled": true,
    "r": 32,
    "alpha": 16,
    "dropout": 0,
    "target_modules": [
      "q_proj",
      "k_proj",
      "v_proj",
      "o_proj",
      "gate_proj",
      "up_proj",
      "down_proj"
    ]
  },

  "distributed_training": {
    "fsdp_config": {
      "enabled": false,
      "sharding_strategy": "FULL_SHARD",
      "mixed_precision": "BF16",
      "activation_checkpointing": true,
      "offload_params": false
    },
    "ddp_find_unused_parameters": false,
    "dataloader_num_workers": 2,
    "ddp_config": {
      "enabled": true,
      "backend": "nccl",
      "find_unused_parameters": false,
      "broadcast_buffers": false,
      "gradient_as_bucket_view": true
    }
  },

  "logging": {
    "logging_steps": 50,
    "log_level": "info"
  },

  "huggingface_hub": {
    "push_to_hub": true,
    "hub_model_id": "phi-4-cognitive-assistant",
    "hub_private_repo": true
  },

  "model_name_or_path": "unsloth/phi-4-unsloth-bnb-4bit",
  "model_revision": "main",
  "use_flash_attention": true,
  "torch_dtype": "bfloat16",
  "bf16": true,
  "fp16": false,

  "hardware": {
    "hardware_name": "4xL4",
    "specs": {
      "gpu_count": 4,
      "gpu_type": "L4",
      "vram_per_gpu": 24,
      "total_vram": 96,
      "vcpu_count": 48,
      "ram": 186
    },
    "hardware_setup": {
      "use_cpu": false,
      "num_gpus": 4,
      "device_map": "auto"
    },
    "training_optimizations": {
      "per_device_batch_size": 16,
      "gradient_accumulation_steps": 3,
      "mixed_precision": "bf16",
      "torch_compile": false,
      "memory_optimizations": {
        "use_gradient_checkpointing": true,
        "use_flash_attention": true
      },
      "multi_gpu_strategy": "ddp"
    },
    "system_settings": {
      "cuda_memory_fraction": 0.75,
      "dataloader_num_workers": 4,
      "dataloader_pin_memory": true
    },
    "memory_breakdown": {
      "model_size": "~3.5GB (pre-quantized 4-bit)",
      "optimizer_states": "~1GB",
      "batch_memory_per_gpu": "~3GB",
      "peak_memory_estimate": "~18GB",
      "safe_headroom": "~6GB"
    },
    "compute_environment": "L4_CLOUD"
  },

  "dataset": {
    "dataset": {
      "name": "George-API/phi4-cognitive-dataset",
      "split": "train"
    },
    "data_formatting": {
      "chat_template": "phi",
      "conversation_structure": {
        "system_identifier": "[RESEARCH INTRODUCTION]",
        "turn_order": ["human", "assistant"]
      },
      "roles": {
        "system": "System: {content}\n\n",
        "human": "Human: {content}\n\n",
        "assistant": "Assistant: {content}\n\n"
      }
    },
    "data_loading": {
      "batch_size": 24,
      "shuffle": false,
      "sequential_processing": true,
      "drop_last": false,
      "num_workers": 4,
      "pin_memory": true,
      "prefetch_factor": 4
    },
    "validation": {
      "log_samples": 3,
      "log_interval": 50,
      "verify_sequence_integrity": true,
      "metrics": ["processed", "skipped", "avg_tokens", "unique_articles"]
    }
  }
}