checkpoints/lang_phase2/hparams.yaml

_target_: salad.models.language_phase2.LangPhase2Model
network:
  _target_: salad.model_components.network.CondDiffNetwork
  input_dim: 512
  residual: true
  context_dim: 784 # concat of 768 lang feat and gaussian.
  context_embedding_dim: 1024
  embedding_dim: 512
  encoder_use_time: false
  encoder_type: transformer
  decoder_type: transformer_encoder
  enc_num_layers: 6
  dec_num_layers: 6
  use_timestep_embedder: true
  timestep_embedder_dim: 128

variance_schedule:
  _target_: salad.model_components.variance_schedule.VarianceSchedule
  num_steps: &time_steps 1000
  beta_1: 1e-4
  beta_T: 0.05
  mode: linear

# optimizer
lr: 1e-4
batch_size: 64

# dataset 
dataset_kwargs:
  data_path: spaghetti_chair_latents.hdf5
  repeat: 1
  data_keys: ["s_j_affine", "g_js_affine"]
  only_easy_context: false
  global_normalization: &normalization false

global_normalization: *normalization
num_timesteps: *time_steps
faster: true
validation_step: 10
no_run_validation: false
spaghetti_tag: "chairs_large" # or airplanes, tables

text_encoder_freeze: false
use_lstm: true
classifier_free_guidance: true
conditioning_dropout_prob: 0.2