Upload config.json with huggingface_hub

config.json

@@ -0,0 +1,60 @@
+{
+  "_name_or_path": "kerzgrr/imagenv1m",
+  "model_type": "imagen-dit",
+  "model_version": "v1-medium",
+  "model_size": "320M",
+  
+  "architecture": {
+    "type": "DiT",
+    "hidden_size": 1024,
+    "depth": 12,
+    "num_heads": 16,
+    "patch_size": 2,
+    "mlp_ratio": 4.0,
+    "context_dim": 1024,
+    "in_channels": 4,
+    "use_rope": true,
+    "use_qk_norm": true,
+    "use_swiglu": true
+  },
+  
+  "text_encoder": {
+    "type": "t5",
+    "model_name": "google/flan-t5-large",
+    "max_length": 256
+  },
+  
+  "vae": {
+    "type": "sdxl",
+    "model_name": "stabilityai/sdxl-vae",
+    "latent_channels": 4,
+    "scaling_factor": 0.13025
+  },
+  
+  "diffusion": {
+    "num_timesteps": 1000,
+    "schedule": "cosine",
+    "prediction_type": "epsilon"
+  },
+  
+  "generation": {
+    "default_steps": 50,
+    "default_cfg_scale": 7.5,
+    "image_size": 256
+  },
+  
+  "training": {
+    "dataset": "UCSC-VLAA/Recap-DataComp-1B",
+    "precision": "bf16",
+    "ema_decay": 0.9999
+  },
+  
+  "model_family": {
+    "v1-nano": {"params": "50M", "hidden_size": 384, "depth": 8, "status": "planned"},
+    "v1-small": {"params": "150M", "hidden_size": 768, "depth": 10, "status": "planned"},
+    "v1-medium": {"params": "320M", "hidden_size": 1024, "depth": 12, "status": "released"},
+    "v1-large": {"params": "700M", "hidden_size": 1536, "depth": 16, "status": "planned"},
+    "v1-xlarge": {"params": "1.5B", "hidden_size": 2048, "depth": 24, "status": "planned"}
+  }
+}
+