Upload folder using huggingface_hub

checkpoint-17000/ema_model/config.json

@@ -0,0 +1,39 @@
+{
+  "_class_name": "MaskGiTUViT_v2",
+  "_version": "0.0.1",
+  "add_cond_embeds": true,
+  "add_micro_cond_embeds": true,
+  "attention_dropout": 0.0,
+  "block_num_heads": 12,
+  "block_out_channels": [
+    768
+  ],
+  "codebook_size": 8192,
+  "cond_embed_dim": 768,
+  "decay": 0.9999,
+  "encoder_hidden_size": 768,
+  "force_down_up_sample": true,
+  "hidden_dropout": 0.0,
+  "hidden_size": 1024,
+  "in_channels": 768,
+  "intermediate_size": 2816,
+  "inv_gamma": 1.0,
+  "layer_norm_eps": 1e-06,
+  "ln_elementwise_affine": true,
+  "mask_token_id": 8255,
+  "micro_cond_embed_dim": 1280,
+  "micro_cond_encode_dim": 256,
+  "min_decay": 0.0,
+  "norm_type": "rmsnorm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 22,
+  "num_res_blocks": 3,
+  "optimization_step": 8000,
+  "power": 0.6666666666666666,
+  "update_after_step": 0,
+  "use_bias": false,
+  "use_ema_warmup": false,
+  "use_fused_mlp": false,
+  "use_fused_residual_norm": false,
+  "vocab_size": 8256
+}

checkpoint-17000/ema_model/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aec69ccd04fba0d4f3eb9e5821acaa0f435118c54918cc42c38a8807f31b0792
+size 2433247453

checkpoint-17000/metadata.json

@@ -0,0 +1 @@
+{"global_step": 17000}

checkpoint-17000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bfe144b120084821b26c0b83f7f96d98d9f6331d06e578c9c8988f6d6fe9f08
+size 4866486533

checkpoint-17000/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b40d6d634d074eb6a20944a25d9a9b516349721258a39c97fa077d84d4abf74b
+size 2433254429

checkpoint-17000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76b347b906ddaec8c955e47f4a42fce503774fa719cbb7c0f1f50a8f69dc604e
+size 14599

checkpoint-17000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:927cb1e3c51f59387f003f7128d398b83c764d5386962b2ff194e10e35375b8a
+size 627

checkpoint-17000/unwrapped_model/config.json

@@ -0,0 +1,32 @@
+{
+  "_class_name": "MaskGiTUViT_v2",
+  "_version": "0.0.1",
+  "add_cond_embeds": true,
+  "add_micro_cond_embeds": true,
+  "attention_dropout": 0.0,
+  "block_num_heads": 12,
+  "block_out_channels": [
+    768
+  ],
+  "codebook_size": 8192,
+  "cond_embed_dim": 768,
+  "encoder_hidden_size": 768,
+  "force_down_up_sample": true,
+  "hidden_dropout": 0.0,
+  "hidden_size": 1024,
+  "in_channels": 768,
+  "intermediate_size": 2816,
+  "layer_norm_eps": 1e-06,
+  "ln_elementwise_affine": true,
+  "mask_token_id": 8255,
+  "micro_cond_embed_dim": 1280,
+  "micro_cond_encode_dim": 256,
+  "norm_type": "rmsnorm",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 22,
+  "num_res_blocks": 3,
+  "use_bias": false,
+  "use_fused_mlp": false,
+  "use_fused_residual_norm": false,
+  "vocab_size": 8256
+}

checkpoint-17000/unwrapped_model/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b40d6d634d074eb6a20944a25d9a9b516349721258a39c97fa077d84d4abf74b
+size 2433254429

config.yaml

@@ -0,0 +1,103 @@
+wandb:
+  entity: r-ap
+  run_id: tx78bpg1
+experiment:
+  name: muse-multi
+  project: muse-prod
+  output_dir: output/
+  max_train_examples: 28500
+  max_eval_examples: 1000
+  save_every: 1000
+  eval_every: 1000
+  generate_every: 400
+  log_every: 50
+  log_grad_norm_every: 100000000
+  resume_from_checkpoint: latest
+  resume_lr_scheduler: true
+  checkpoints_total_limit: 4
+  logging_dir: output/logs
+model:
+  vq_model:
+    type: vqgan
+  text_encoder:
+    type: clip
+  transformer:
+    vocab_size: 8256
+    hidden_size: 1024
+    intermediate_size: 2816
+    num_hidden_layers: 22
+    num_attention_heads: 16
+    in_channels: 768
+    block_out_channels:
+    - 768
+    block_has_attention:
+    - true
+    block_num_heads: 12
+    num_res_blocks: 3
+    res_ffn_factor: 4
+    patch_size: 1
+    encoder_hidden_size: 768
+    add_cross_attention: true
+    project_encoder_hidden_states: true
+    codebook_size: 8192
+    num_vq_tokens: 256
+    initializer_range: 0.02
+    norm_type: rmsnorm
+    layer_norm_eps: 1.0e-06
+    ln_elementwise_affine: true
+    use_encoder_layernorm: false
+    use_bias: false
+    hidden_dropout: 0.0
+    attention_dropout: 0.0
+    use_codebook_size_for_output: true
+    use_empty_embeds_for_uncond: true
+    add_cond_embeds: true
+    cond_embed_dim: 768
+    add_micro_cond_embeds: true
+    micro_cond_encode_dim: 256
+    micro_cond_embed_dim: 1280
+    force_down_up_sample: true
+  architecture: uvit
+  enable_xformers_memory_efficient_attention: true
+dataset:
+  preprocessing:
+    max_seq_length: 77
+    resolution: 256
+optimizer:
+  name: adamw
+  params:
+    learning_rate: 0.0001
+    scale_lr: false
+    beta1: 0.9
+    beta2: 0.999
+    weight_decay: 0.01
+    epsilon: 1.0e-08
+lr_scheduler:
+  scheduler: constant_with_warmup
+  params:
+    learning_rate: ${optimizer.params.learning_rate}
+    warmup_steps: 100
+training:
+  gradient_accumulation_steps: 1
+  batch_size: 20
+  mixed_precision: 'no'
+  enable_tf32: true
+  use_ema: true
+  ema_decay: 0.9999
+  ema_update_after_step: 0
+  ema_update_every: 1
+  seed: 13399
+  max_train_steps: 20000
+  overfit_one_batch: false
+  cond_dropout_prob: 0.1
+  min_masking_rate: 0.0
+  label_smoothing: 0.1
+  max_grad_norm: null
+  guidance_scale: 8
+  generation_timesteps: 16
+  use_soft_code_target: false
+  use_stochastic_code: false
+  soft_code_temp: 1.0
+  mask_schedule: cosine
+  mask_contiguous_region_prob: 0.15
+config: configs/segmentation.yaml