Upload model_config.json with huggingface_hub

model_config.json

@@ -0,0 +1,131 @@
+{
+    "model_type": "diffusion_cond",
+    "sample_size": 882000,
+    "sample_rate": 44100,
+    "audio_channels": 2,
+    "model": {
+        "pretransform": {
+            "type": "autoencoder",
+            "iterate_batch": true,
+            "config": {
+                "encoder": {
+                    "type": "oobleck",
+                    "requires_grad": false,
+                    "config": {
+                        "in_channels": 2,
+                        "channels": 128,
+                        "c_mults": [1, 2, 4, 8, 16],
+                        "strides": [2, 4, 4, 8, 8],
+                        "latent_dim": 128,
+                        "use_snake": true
+                    }
+                },
+                "decoder": {
+                    "type": "oobleck",
+                    "config": {
+                        "out_channels": 2,
+                        "channels": 128,
+                        "c_mults": [1, 2, 4, 8, 16],
+                        "strides": [2, 4, 4, 8, 8],
+                        "latent_dim": 64,
+                        "use_snake": true,
+                        "final_tanh": false
+                    }
+                },
+                "bottleneck": {
+                    "type": "vae"
+                },
+                "latent_dim": 64,
+                "downsampling_ratio": 2048,
+                "io_channels": 2
+            }
+        },
+        "conditioning": {
+            "configs": [
+                {
+                    "id": "prompt",
+                    "type": "t5",
+                    "config": {
+                        "t5_model_name": "t5-base",
+                        "max_length": 128
+                    }
+                },
+                {
+                    "id": "seconds_start",
+                    "type": "number",
+                    "config": {
+                        "min_val": 0,
+                        "max_val": 512
+                    }
+                },
+                {
+                    "id": "seconds_total",
+                    "type": "number",
+                    "config": {
+                        "min_val": 0,
+                        "max_val": 512
+                    }
+                }
+            ],
+            "cond_dim": 768
+        },
+        "diffusion": {
+            "cross_attention_cond_ids": ["prompt", "seconds_start", "seconds_total"],
+            "global_cond_ids": ["seconds_start", "seconds_total"],
+            "type": "dit",
+            "config": {
+                "io_channels": 64,
+                "embed_dim": 1536,
+                "depth": 24,
+                "num_heads": 24,
+                "cond_token_dim": 768,
+                "global_cond_dim": 1536,
+                "project_cond_tokens": false,
+                "transformer_type": "continuous_transformer"
+            }
+        },
+        "io_channels": 64
+    },
+    "training": {
+        "use_ema": true,
+        "log_loss_info": false,
+        "optimizer_configs": {
+            "diffusion": {
+                "optimizer": {
+                    "type": "AdamW",
+                    "config": {
+                        "lr": 5e-5,
+                        "betas": [0.9, 0.999],
+                        "weight_decay": 1e-3
+                    }
+                },
+                "scheduler": {
+                    "type": "InverseLR",
+                    "config": {
+                        "inv_gamma": 1000000,
+                        "power": 0.5,
+                        "warmup": 0.99
+                    }
+                }
+            }
+        },
+        "demo": {
+            "demo_every": 14784,
+            "demo_steps": 250,
+            "num_demos": 10,
+            "demo_cond": [
+                {"prompt": "Guitar, Steel Guitar, Mids, Upper Mids, Highs, Pluck, Bright, Clean, simple melody, D minor, 100 BPM, 8 bars", "seconds_start": 0, "seconds_total": 19},
+                {"prompt": "Guitar, Steel Guitar, Mids, Upper Mids, Highs, Pluck, Bright, Clean, simple melody, D minor, 100 BPM, 8 bars", "seconds_start": 0, "seconds_total": 19},
+                {"prompt": "Bowed Strings, Violin, Cello, Mids, Upper Mids, Staccato, Tight, Focused, Smooth, Present, Near, Ensemble, Wet, catchy dance chord progression, with top dance melody, C# minor, 128 BPM, 8 bars", "seconds_start": 0, "seconds_total": 15},
+                {"prompt": "Vocal, Synthetic, Bass, Mids, Upper Mids, Highs, Digital, Retro, Synthetic Vox, D major, 140 BPM, 8 bars", "seconds_start": 0, "seconds_total": 14},
+                {"prompt": "Keys, Grand Piano, Highs, Full, Subdued, Wet, Medium Reverb, Medium Phaser, medium speed, off beat, repeating, melody, C minor, 128 BPM, 8 bars", "seconds_start": 0, "seconds_total": 15},
+                {"prompt": "Bass, FM Bass, Sub Bass, Bass, Upper Mids, Highs, Thick, Clean, Pitch Bend, Wet, Medium Delay, Medium Reverb, Low Distortion, Phaser, 4 bars, 150 BPM, D minor", "seconds_start": 0, "seconds_total": 6},
+                {"prompt": "Wind, World Winds, Flute, Airy, Hollow,Bb major, 110 BPM, 4 bars", "seconds_start": 0, "seconds_total": 8},
+                {"prompt": "Bass, Wavetable Bass, Sub Bass, Bass, Upper Mids, Highs, Acid, 8 bars, 128 BPM, E minor", "seconds_start": 0, "seconds_total": 15},
+                {"prompt": "Ocarina, Formant Vocal, Warm, Rich, Clean, Medium Reverb F minor, 128 BPM, 8 bars", "seconds_start": 0, "seconds_total": 15},
+                {"prompt": "Ocarina, Formant Vocal, Warm, Rich, Clean, Medium Reverb F minor, 128 BPM, 8 bars", "seconds_start": 0, "seconds_total": 15}
+            ],
+            "demo_cfg_scales": [7]
+        }
+    }
+}