Nateraw's workspace
Runs 
1
State
Notes
User
Tags
Created
Runtime
Sweep
attribute_dropout
autocast
autocast_dtype
benchmark_no_load
cache
channels
checkpoint
classifier_free_guidance
codebooks_pattern
compression_model_checkpoint
conditioners
continue_from
dataset
datasource
deadlock
device
dora
dtype
efficient_attention_backend
evaluate
execute_inplace
fsdp
fuser
generate
interleave_stereo_codebooks
lm_model
logging
metrics
mp_start_method
num_threads
optim
profiler
sample_rate
schedule
seed
show
slurm
solver
tensorboard
tokens
transformer_lm
wandb
ce
ce_q1
Finished
-
nateraw
5h 11m 8s
-
{}
true
float16
false
{'path': None, 'write': False, 'write_shard': 0, 'write_num_shards': 1}
2
{'save_last': True, 'save_every': 50, 'keep_last': 10, 'keep_every_states': None}
{'training_dropout': 0.3, 'inference_coef': 3.0}
{'modeling': 'delay', 'delay': {'delays': [0, 1, 2, 3], 'flatten_first': 0, 'empty_initial': 0}, 'unroll': {'flattening': [0, 1, 2, 3], 'delays': [0, 0, 0, 0]}, 'music_lm': {'group_by': 2}, 'coarse_first': {'delays': [0, 0, 0]}}
//pretrained/facebook/encodec_32khz
{'description': {'model': 't5', 't5': {'name': 't5-base', 'finetune': False, 'word_dropout': 0.3, 'normalize_text': False}}}
facebook/musicgen-stereo-melody-large
{'batch_size': 5, 'num_workers': 124, 'segment_duration': 15, 'num_samples': None, 'return_info': True, 'shuffle': False, 'sample_on_duration': False, 'sample_on_weight': False, 'min_segment_ratio': 0.8, 'train': {'num_samples': 1000000, 'shuffle': True, 'shuffle_seed': 0, 'permutation_on_files': False, 'merge_text_p': 0.25, 'drop_desc_p': 0.5, 'drop_other_p': 0.5}, 'valid': {'num_samples': 10000}, 'evaluate': {'num_samples': 10000}, 'generate': {'num_samples': 50, 'return_info': True}}
{'max_sample_rate': 32000, 'max_channels': 2, 'train': 'egs/splice_v2', 'valid': 'egs/splice_v2', 'evaluate': 'egs/splice_v2', 'generate': 'egs/splice_v2'}
{'use': True, 'timeout': 600}
cuda
{'dir': '/checkpoint/${oc.env:USER}/experiments/audiocraft/outputs', 'exclude': ['device', 'wandb.*', 'tensorboard.*', 'logging.*', 'dataset.num_workers', 'eval.num_workers', 'special.*', 'metrics.visqol.bin', 'metrics.fad.bin', 'execute_only', 'execute_best', 'generate.every', 'optim.eager_sync', 'profiler.*', 'deadlock.*', 'efficient_attention_backend', 'num_threads', 'mp_start_method'], 'use_rendezvous': False, 'git_save': True}
float32
torch
{'every': 25, 'num_workers': 5, 'truncate_audio': None, 'fixed_generation_duration': None, 'metrics': {'base': False, 'fad': False, 'kld': False, 'text_consistency': False, 'chroma_cosine': False}}
false
{'use': True, 'param_dtype': 'float16', 'reduce_dtype': 'float32', 'buffer_dtype': 'float32', 'sharding_strategy': 'shard_grad_op', 'per_block': True}
{'cross_attention_pos_emb': False, 'cross_attention_pos_emb_scale': 1, 'sum': [], 'prepend': [], 'cross': ['description'], 'input_interpolate': []}
{'every': 25, 'path': 'samples', 'audio': {'format': 'wav', 'strategy': 'loudness', 'sample_rate': '${sample_rate}', 'loudness_headroom_db': 14}, 'lm': {'use_sampling': True, 'temp': 1.0, 'top_k': 250, 'top_p': 0.0, 'prompted_samples': True, 'unprompted_samples': True, 'gen_gt_samples': False, 'prompt_duration': None, 'gen_duration': None, 'remove_prompts': False}, 'num_workers': 5}
{'use': True, 'per_timestep': False}
transformer_lm
{'level': 'INFO', 'log_updates': 10, 'log_tensorboard': True, 'log_wandb': True}
{'fad': {'use_gt': False, 'model': 'tf', 'tf': {'bin': None, 'model_path': '//reference/fad/vggish_model.ckpt'}}, 'kld': {'use_gt': False, 'model': 'passt', 'passt': {'pretrained_length': 20}}, 'text_consistency': {'use_gt': False, 'model': 'clap', 'clap': {'model_path': '//reference/clap/music_audioset_epoch_15_esc_90.14.pt', 'model_arch': 'HTSAT-base', 'enable_fusion': False}}, 'chroma_cosine': {'use_gt': False, 'model': 'chroma_base', 'chroma_base': {'sample_rate': '${sample_rate}', 'n_chroma': 12, 'radix2_exp': 14, 'argmax': True}}}
forkserver
1
{'epochs': 500, 'updates_per_epoch': 2000, 'lr': 1, 'optimizer': 'dadam', 'adam': {'betas': [0.9, 0.95], 'weight_decay': 0.1, 'eps': 1e-08}, 'ema': {'use': True, 'updates': 10, 'device': 'cuda', 'decay': 0.99}, 'max_norm': 1.0, 'eager_sync': True}
{'enabled': False}
32000
{'lr_scheduler': 'cosine', 'step': {'step_size': None, 'gamma': None}, 'exponential': {'lr_decay': None}, 'cosine': {'warmup': 200, 'lr_min_ratio': 0.0, 'cycle_length': 1.0}, 'polynomial_decay': {'warmup': None, 'zero_lr_warmup_steps': 0, 'end_lr': 0.0, 'power': 1}, 'inverse_sqrt': {'warmup': None, 'warmup_init_lr': 0.0}, 'linear_warmup': {'warmup': None, 'warmup_init_lr': 0.0}}
2036
false
{'gpus': 4, 'mem_per_gpu': 40, 'time': 3600, 'constraint': None, 'partition': None, 'comment': None, 'setup': [], 'exclude': ''}
musicgen
{'with_media_logging': False, 'name': None, 'sub_dir': None}
{'padding_with_special_token': False}
{'dim': 1536, 'num_heads': 24, 'num_layers': 48, 'hidden_scale': 4, 'n_q': 4, 'card': 2048, 'dropout': 0.0, 'emb_lr': None, 'activation': 'gelu', 'norm_first': True, 'bias_ff': False, 'bias_attn': False, 'bias_proj': False, 'past_context': None, 'causal': True, 'custom': False, 'memory_efficient': True, 'attention_as_float32': False, 'layer_scale': None, 'positional_embedding': 'sin', 'xpos': False, 'checkpointing': 'none', 'weight_init': 'gaussian', 'depthwise_init': 'current', 'zero_bias_init': True, 'norm': 'layer_norm', 'cross_attention': False, 'qk_layer_norm': False, 'qk_layer_norm_cross': False, 'attention_dropout': None, 'kv_repeat': 1, 'two_step_cfg': False}
{'with_media_logging': True, 'project': 'audiocraft-48k-stereo-lightning', 'name': 'splice_v2-musicgen-32k-stereo-large-ft-facebook/musicgen-stereo-melody-large', 'group': None}
2.69584
2.01217
1-1
of 1