Upload qwenvl2/checkpoint-309000/model_ema/config.json with huggingface_hub

qwenvl2/checkpoint-309000/model_ema/config.json

@@ -0,0 +1,53 @@
+{
+  "_class_name": "OpenSoraT2I",
+  "_diffusers_version": "0.30.2",
+  "activation_fn": "gelu-approximate",
+  "attention_bias": true,
+  "attention_head_dim": 64,
+  "caption_channels": 3584,
+  "caption_channels_2": null,
+  "conv_ffn": false,
+  "conv_out": false,
+  "decay": 0.999,
+  "deepnorm": false,
+  "dropout": 0.0,
+  "explicit_uniform_rope": false,
+  "in_channels": 32,
+  "interpolation_scale_h": 1.0,
+  "interpolation_scale_t": 1.0,
+  "interpolation_scale_w": 1.0,
+  "inv_gamma": 1.0,
+  "layerwise_text_mlp": false,
+  "min_decay": 0.0,
+  "norm_cls": "fp32_layer_norm",
+  "norm_elementwise_affine": true,
+  "norm_eps": 1e-06,
+  "norm_skip": false,
+  "num_attention_heads": 24,
+  "num_layers": [
+    12,
+    1,
+    12
+  ],
+  "optimization_step": 309000,
+  "out_channels": 32,
+  "patch_size": 2,
+  "patch_size_t": 1,
+  "pooled_projection_dim": 0,
+  "power": 0.6666666666666666,
+  "prenorm_num": 1000,
+  "sample_size_h": 16,
+  "sample_size_t": 1,
+  "sample_size_w": 16,
+  "sandwich_norm": false,
+  "shrink": false,
+  "skip_connection": false,
+  "sparse1d": false,
+  "sparse_n": 2,
+  "time_as_text_token": false,
+  "time_as_x_token": false,
+  "timestep_embed_dim": 512,
+  "update_after_step": 0,
+  "use_ema_warmup": false,
+  "use_text_dim": false
+}