Add files using upload-large-folder tool

architecture.txt

@@ -0,0 +1,10 @@
+model_tag: moe_full
+task: peg-insert-side-v3
+config: experiments/moe_full_unfreeze.yaml
+action_head_type: moe
+router_condition: action_input
+freeze_vision: true
+freeze_text: true
+unfreeze_vision_last_n_layers: 2
+unfreeze_text_last_n_layers: 2
+act_chunk_size: n/a

moe_router_entropy.csv

@@ -0,0 +1,17 @@
+epoch,task_name,mean_router_entropy,sample_count
+1,short_metaworld,1.0972516559274759,4250
+2,short_metaworld,0.868278188122546,4250
+3,short_metaworld,0.8644592178498998,4250
+4,short_metaworld,0.8801690644621849,4250
+5,short_metaworld,0.9024338802160585,4250
+6,short_metaworld,0.9231361929274657,4250
+7,short_metaworld,0.9505329890706959,4250
+8,short_metaworld,0.9680941077444484,4250
+9,short_metaworld,0.9897653412012493,4250
+10,short_metaworld,1.0097954931697424,4250
+11,short_metaworld,1.019337093510172,4250
+12,short_metaworld,1.0326422810493148,4250
+13,short_metaworld,1.0416921249838436,4250
+14,short_metaworld,1.0466027086096652,4250
+15,short_metaworld,1.0440011160680476,4250
+16,short_metaworld,1.0458287356697462,4250

moe_router_weights.csv

@@ -0,0 +1,65 @@
+epoch,task_name,expert_idx,mean_router_weight,sample_count
+1,short_metaworld,0,0.2524724304676056,4250
+1,short_metaworld,1,0.2697482109069824,4250
+1,short_metaworld,2,0.26585105061531067,4250
+1,short_metaworld,3,0.21192866563796997,4250
+2,short_metaworld,0,0.3143608272075653,4250
+2,short_metaworld,1,0.3268682062625885,4250
+2,short_metaworld,2,0.21709245443344116,4250
+2,short_metaworld,3,0.14167837798595428,4250
+3,short_metaworld,0,0.3089982569217682,4250
+3,short_metaworld,1,0.3164787292480469,4250
+3,short_metaworld,2,0.2356213927268982,4250
+3,short_metaworld,3,0.13890130817890167,4250
+4,short_metaworld,0,0.29757270216941833,4250
+4,short_metaworld,1,0.30781999230384827,4250
+4,short_metaworld,2,0.2540871202945709,4250
+4,short_metaworld,3,0.1405191272497177,4250
+5,short_metaworld,0,0.28526824712753296,4250
+5,short_metaworld,1,0.29758328199386597,4250
+5,short_metaworld,2,0.27030232548713684,4250
+5,short_metaworld,3,0.14684562385082245,4250
+6,short_metaworld,0,0.27950519323349,4250
+6,short_metaworld,1,0.289709210395813,4250
+6,short_metaworld,2,0.27237799763679504,4250
+6,short_metaworld,3,0.15840782225131989,4250
+7,short_metaworld,0,0.26899853348731995,4250
+7,short_metaworld,1,0.27762043476104736,4250
+7,short_metaworld,2,0.2824629247188568,4250
+7,short_metaworld,3,0.17091910541057587,4250
+8,short_metaworld,0,0.26036784052848816,4250
+8,short_metaworld,1,0.26750805974006653,4250
+8,short_metaworld,2,0.2912062704563141,4250
+8,short_metaworld,3,0.18091799318790436,4250
+9,short_metaworld,0,0.2523835301399231,4250
+9,short_metaworld,1,0.2574305534362793,4250
+9,short_metaworld,2,0.2924705445766449,4250
+9,short_metaworld,3,0.19771520793437958,4250
+10,short_metaworld,0,0.23678657412528992,4250
+10,short_metaworld,1,0.2441885620355606,4250
+10,short_metaworld,2,0.29841580986976624,4250
+10,short_metaworld,3,0.22060911357402802,4250
+11,short_metaworld,0,0.22633861005306244,4250
+11,short_metaworld,1,0.23437286913394928,4250
+11,short_metaworld,2,0.3036373257637024,4250
+11,short_metaworld,3,0.23565179109573364,4250
+12,short_metaworld,0,0.220066636800766,4250
+12,short_metaworld,1,0.22533780336380005,4250
+12,short_metaworld,2,0.2991962134838104,4250
+12,short_metaworld,3,0.25539910793304443,4250
+13,short_metaworld,0,0.21208599209785461,4250
+13,short_metaworld,1,0.21645447611808777,4250
+13,short_metaworld,2,0.2984868884086609,4250
+13,short_metaworld,3,0.2729724049568176,4250
+14,short_metaworld,0,0.21153134107589722,4250
+14,short_metaworld,1,0.2144383043050766,4250
+14,short_metaworld,2,0.2981346845626831,4250
+14,short_metaworld,3,0.275896281003952,4250
+15,short_metaworld,0,0.20906934142112732,4250
+15,short_metaworld,1,0.21298854053020477,4250
+15,short_metaworld,2,0.29622671008110046,4250
+15,short_metaworld,3,0.2817155122756958,4250
+16,short_metaworld,0,0.2072872817516327,4250
+16,short_metaworld,1,0.21117538213729858,4250
+16,short_metaworld,2,0.2976193130016327,4250
+16,short_metaworld,3,0.2839184105396271,4250

tensorboard/events.out.tfevents.1776852769.praise-5080.2844183.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9c96de0901951335e66869ff79d7ccf87555bae5029154ea4971f1e68d10ad4
+size 399563

train_config.json

@@ -0,0 +1,59 @@
+{
+  "dataset_type": "short_metaworld",
+  "data_root": "data/short-metaworld-vla",
+  "train_jsonl": "",
+  "val_jsonl": "",
+  "val_ratio": 0.15,
+  "vision_model_name": "google/siglip2-base-patch16-224",
+  "text_model_name": "google/siglip2-base-patch16-224",
+  "separate_backbones": false,
+  "image_size": 224,
+  "freeze_vision": true,
+  "freeze_text": true,
+  "unfreeze_vision_last_n_layers": 2,
+  "unfreeze_text_last_n_layers": 2,
+  "fusion_type": "cross_attn",
+  "proj_dim": 512,
+  "fusion_hidden_dim": 1024,
+  "fusion_out_dim": 512,
+  "fusion_num_layers": 3,
+  "fusion_num_heads": 8,
+  "fusion_dropout": 0.1,
+  "normalize_embeddings": true,
+  "action_head_type": "moe",
+  "action_mlp_hidden_dim": 256,
+  "action_mlp_layers": 2,
+  "action_mlp_dropout": 0.1,
+  "moe_num_experts": 4,
+  "moe_hidden_dim": 512,
+  "moe_load_balance_weight": 0.001,
+  "router_condition": "action_input",
+  "act_chunk_size": 8,
+  "act_hidden_dim": 512,
+  "act_num_layers": 2,
+  "act_dropout": 0.1,
+  "use_geometry_features": true,
+  "geometry_dim": 6,
+  "temporal_context": 4,
+  "action_dim": 4,
+  "num_workers": 8,
+  "normalize_action_targets": true,
+  "action_norm_eps": 1e-06,
+  "learnable_action_scale": true,
+  "action_scale_init": 1.0,
+  "loss_type": "huber",
+  "huber_delta": 0.5,
+  "epochs": 80,
+  "batch_size": 32,
+  "grad_accum_steps": 2,
+  "learning_rate": 5e-05,
+  "weight_decay": 0.01,
+  "grad_clip_norm": 1.0,
+  "use_fp16": true,
+  "early_stopping_patience": 8,
+  "early_stopping_min_delta": 0.0005,
+  "seed": 42,
+  "device": "cuda",
+  "out_dir": "checkpoints_stage2_moe_full_peg_only",
+  "save_best_by_val": true
+}