Hugging Face's logo

arcinstitute
/
ST-Parse

Safetensors
Model card Files
xet
Community
2
ST-Parse
File size: 2,315 Bytes 
423a5ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
data:
  name: PerturbationDataModule
  kwargs:
    toml_config_path: /large_storage/ctc/ML/state_sets/parse/donor.toml
    embed_key: X_hvg
    output_space: gene
    pert_rep: onehot
    basal_rep: sample
    num_workers: 24
    pin_memory: true
    n_basal_samples: 1
    basal_mapping_strategy: random
    should_yield_control_cells: true
    batch_col: cell_type
    pert_col: cytokine
    cell_type_key: donor
    control_pert: PBS
    map_controls: true
    perturbation_features_file: null
    store_raw_basal: false
    int_counts: false
  output_dir: null
  debug: true
model:
  name: PertSets
  checkpoint: null
  device: cuda
  kwargs:
    cell_set_len: 512
    blur: 0.05
    hidden_dim: 1440
    loss: energy
    confidence_head: false
    n_encoder_layers: 4
    n_decoder_layers: 4
    predict_residual: true
    softplus: true
    freeze_pert: false
    transformer_decoder: false
    finetune_vci_decoder: false
    residual_decoder: false
    batch_encoder: true
    nb_decoder: false
    mask_attn: false
    use_effect_gating_token: false
    distributional_loss: energy
    init_from: /large_storage/ctc/userspace/aadduri/preprint_parse/parse_donor_tahoe_best_cs512_2000_hvg/parse_donor_tahoe_best_2000/checkpoints/step=100000.ckpt
    transformer_backbone_key: llama
    transformer_backbone_kwargs:
      max_position_embeddings: 512
      hidden_size: 1440
      intermediate_size: 4416
      num_hidden_layers: 4
      num_attention_heads: 12
      num_key_value_heads: 12
      head_dim: 120
      use_cache: false
      attention_dropout: 0.0
      hidden_dropout: 0.0
      layer_norm_eps: 1.0e-06
      pad_token_id: 0
      bos_token_id: 1
      eos_token_id: 2
      tie_word_embeddings: false
      rotary_dim: 0
      use_rotary_embeddings: false
training:
  wandb_track: false
  weight_decay: 0.0005
  batch_size: 32
  lr: 0.0001
  max_steps: 10
  train_seed: 42
  val_freq: 100
  ckpt_every_n_steps: 1
  gradient_clip_val: 10
  loss_fn: mse
wandb:
  entity: arcinstitute
  project: vci1
  local_wandb_dir: /large_storage/ctc/userspace/aadduri/wandb_dir/vci1
  tags:
  - parse_fig2_final_donor
name: cs512
output_dir: /large_storage/ctc/userspace/aadduri/preprint/parse_fig2_release
use_wandb: true
overwrite: false
return_adatas: false
pred_adata_path: null
true_adata_path: null