Upload folder using huggingface_hub

lam/config.json

@@ -12,7 +12,7 @@
   ],
   "decoder_attention_head_dim": 64,
   "decoder_attn_implementation": "flash_attention_2",
-  "decoder_encoder_hidden_dim": 3,
+  "decoder_encoder_hidden_dim": 4,
   "decoder_eps": 1e-06,
   "decoder_ffn_dim": 768,
   "decoder_freq_dim": 64,
@@ -32,12 +32,14 @@
   "fsq_levels": [
     8,
     5,
+    5,
     5
   ],
   "initializer_range": 0.02,
+  "is_action_discrete": false,
   "is_diffusion": true,
-  "latent_channels": 3,
-  "max_tokens": 128,
+  "latent_channels": 4,
+  "max_tokens": 64,
   "min_tokens": 1,
   "model_type": "lam",
   "null_latent": 0,
@@ -71,5 +73,5 @@
     "with_cp": false
   },
   "videomae_from_pretrained": null,
-  "vocab_size": 200
+  "vocab_size": 1000
 }

lam/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b9cd4d00c3766e9b28f89158ade6ffe30d64eef6ba8929f96398475679416c40
-size 23418680
+oid sha256:d8bc80515fb9cf54863015c22ac0e8e6a433cbe8be5e4e9e8a8d320c397dca20
+size 23288380

policy/config.json

@@ -1,21 +1,21 @@
 {
-  "action_latent_dim": null,
-  "action_seq_len": 128,
-  "action_start_token_id": 203,
-  "action_vocab_size": 200,
+  "action_seq_len": 64,
+  "action_start_token_id": 1003,
+  "action_vocab_size": 1000,
   "architectures": [
     "PolicyQwen3ForConditionalGeneration"
   ],
   "attention_bias": false,
   "attention_dropout": 0.0,
   "dtype": "bfloat16",
-  "eos_token_id": 204,
-  "frame_stride": 3,
+  "eos_token_id": 1004,
+  "frame_stride": 1,
   "head_dim": 64,
   "hidden_act": "silu",
   "hidden_size": 256,
-  "image_token_id": 202,
+  "image_token_id": 1002,
   "initializer_range": 0.02,
+  "input_frame_stride": 8,
   "intermediate_size": 1024,
   "layer_types": [
     "full_attention",
@@ -31,17 +31,16 @@
     "full_attention",
     "full_attention"
   ],
-  "max_position_embeddings": 2048,
+  "max_position_embeddings": 8192,
   "max_window_layers": 28,
   "model_type": "policy_qwen3",
   "num_attention_heads": 4,
-  "num_frames": 2,
+  "num_frames": 10,
   "num_hidden_layers": 12,
   "num_key_value_heads": 4,
   "pad_token_id": 0,
   "policy_image_height": 64,
   "policy_image_width": 64,
-  "predict_tokens": true,
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
   "rope_theta": 10000.0,
@@ -50,10 +49,10 @@
   "transformers_version": "4.57.1",
   "use_cache": false,
   "use_sliding_window": false,
-  "vision_end_token_id": 201,
+  "vision_end_token_id": 1001,
   "vision_in_channels": 3,
   "vision_merge_size": 1,
-  "vision_patch_size": 4,
-  "vision_start_token_id": 200,
-  "vocab_size": 205
+  "vision_patch_size": 8,
+  "vision_start_token_id": 1000,
+  "vocab_size": 1005
 }

policy/generation_config.json

@@ -1,6 +1,6 @@
 {
   "_from_model_config": true,
-  "eos_token_id": 204,
+  "eos_token_id": 1004,
   "pad_token_id": 0,
   "transformers_version": "4.57.1",
   "use_cache": false

policy/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6cf1309a302321033f67c7e02769544951c7c3da13d96c08c38513a9fc857dae
-size 25326488
+oid sha256:56c0bb93a2d7e4012eeb5f57a7b6a9512fcf68630f835ee05b1be3083a0a54e8
+size 25809824

policy_processor/preprocessor_config.json

@@ -17,7 +17,7 @@
   "max_pixels": 1003520,
   "merge_size": 1,
   "min_pixels": 3136,
-  "patch_size": 4,
+  "patch_size": 8,
   "resample": 3,
   "rescale_factor": 0.00392156862745098,
   "size": {