Hugging Face's logo

fineinstructions
/
pretraining_experiments

Safetensors
Model card Files
xet
Community
pretraining_experiments
File size: 2,289 Bytes 
cb16670
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
name: fineinstructions_ipt_synthetic_all_exp
dump_dir: /fsx/craffel/fineinstructions/pretraining/ipt_synthetic_all_exp/
seed: 777
grad_acc_steps: 8
gc_collect_freq: 1000
probe_freq: null
steps: 22000
data:
  root_dir: /scratch/craffel/lingua/data/fineinstructions/
  sources:
    ipt_synthetic_all_exp: 1.0
  batch_size: 4
  seq_len: 4096
  n_views: 2
  seed: 42
  add_bos: true
  add_eos: true
  load_async: true
  prefetch_size: 1024
  tokenizer:
    name: tiktoken
    path: /fsx/craffel/lingua/tokenizers/llama3.model
    n_words: null
optim:
  lr: 0.001
  weight_decay: 0.1
  epsilon: 1.0e-08
  beta1: 0.9
  beta2: 0.95
  clip: 1.0
  scheduler: cosine
  warmup: 2000
  lr_min_ratio: 1.0e-06
  cycle_length: 1.0
  cosine_theta: 1.0
  annealing_step: 1000
  decay_fraction: 0.1
  exp_factor: 0.5
model:
  dim: 2048
  n_layers: 25
  head_dim: null
  n_heads: 16
  n_kv_heads: null
  ffn_dim_multiplier: null
  multiple_of: 256
  norm_eps: 1.0e-05
  rope_theta: 10000.0
  init_base_std: null
  init_std_factor: disabled
  max_seqlen: 4096
  seed: 42
  vocab_size: 128256
  weight_tying: false
  sliding_window: null
distributed:
  dp_shard: 1
  dp_replicate: 8
  tp_size: 1
  selective_activation_checkpointing: false
  compile: true
  fsdp_type: full_shard
  model_dtype: bf16
  float8_recipe: null
  float8_filter: layers\.[0-9]+\.
  matmul_allow_tf32: false
  detect_anomaly: false
  compile_cache_size_limit: 8
  spawn_method: forkserver
env:
  MKL_SERVICE_FORCE_INTEL: GNU
  OMP_NUM_THREADS: '1'
  MKL_NUM_THREADS: '1'
  ENABLE_INTRA_NODE_COMM: '1'
  TORCH_NCCL_AVOID_RECORD_STREAMS: '1'
  NCCL_IB_TIMEOUT: '22'
  NCCL_DEBUG: INFO
  TORCH_NCCL_ASYNC_ERROR_HANDLING: '1'
checkpoint:
  dump:
    every: 2000
    keep: -1
  eval:
    every: 2000
    keep: -1
  path: /fsx/craffel/fineinstructions/pretraining/ipt_synthetic_all_exp/checkpoints
  init_ckpt_path: null
  load_init_optimizer_state: false
  save_init_ckpt: false
profiling:
  run: true
  trace_folder: profiling
  mem_warmup: 0
  mem_steps: 4
  profile_warmup: 100
  profile_steps: 4
logging:
  freq: 1
  acc_freq: null
  wandb: null
async_eval_gpus: 8
eval:
  harness:
    tasks:
    - hellaswag
    - mmlu
    - commonsense_qa
    - sciq
    confirm_run_unsafe_code: true
  generator:
    max_tokens: 8192
    dtype: bf16