Upload config.json

config.json

@@ -0,0 +1,43 @@
+{
+  "_gradient_checkpointing": true,
+  "_name_or_path": "Yysrc/Mantis-Base",
+  "action_dim": 7,
+  "action_model_type": "DiT-L",
+  "architectures": [
+    "Mantis"
+  ],
+  "attn_implementation": null,
+  "connector_num_hidden_layers": 12,
+  "diffusion_model_id": "Efficient-Large-Model/Sana_600M_512px_diffusers",
+  "future_action_window_size": 4,
+  "in_channels": 32,
+  "input_size": 16,
+  "loss_type": "flow",
+  "max_input_text_tokens": 256,
+  "max_timestep_gap": 6,
+  "mllm_id": "Qwen/Qwen2.5-VL-3B-Instruct",
+  "model_type": "mantis",
+  "modules_to_freeze": [
+    "vae",
+    "model.mllm_backbone.visual"
+  ],
+  "modules_to_unfreeze": [
+    "model.mllm_backbone.model.embed_tokens",
+    "model.policy_head",
+    "model.transformer",
+    "model.connector",
+    "model.mllm_backbone.model"
+  ],
+  "noise_scheduler_id": "Efficient-Large-Model/Sana_600M_512px_diffusers",
+  "num_actqueries": 6,
+  "num_gapqueries": 3,
+  "num_metaqueries": 9,
+  "past_action_window_size": 0,
+  "scheduler_id": "Efficient-Large-Model/Sana_600M_512px_diffusers",
+  "system_prompt": "You will be provided with an image observation and a corresponding instruction.",
+  "torch_dtype": "float32",
+  "training_mode": "image_action_language",
+  "transformers_version": "4.49.0",
+  "vae_downsample_f": 32,
+  "vae_id": "Efficient-Large-Model/Sana_600M_512px_diffusers"
+}