Add CodeLeWM checkpoint artifact codelewm-action-use-retrieval-20260520-7895d18

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/checkpoints/checkpoint.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cb4daf1500495579f5c59cc9fd8aa39f5f70e88f55c0c121320d023b43ddeda
+size 351495901

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/checkpoints/checkpoint.pt.manifest.json

@@ -0,0 +1,14 @@
+{
+  "checkpoint_path": "checkpoint.pt",
+  "checkpoint_sha256": "0cb4daf1500495579f5c59cc9fd8aa39f5f70e88f55c0c121320d023b43ddeda",
+  "metadata": {
+    "action_view": "text",
+    "config_hash": "9ac2ec1d60645bf2eb351d4799f5866533a26a39d7b3dec8dc064988c19c8cf8",
+    "latent_dim": 256,
+    "model_class": "TorchCodeTransitionModel",
+    "record_schema_version": "codelewm.transition.v1",
+    "schema_version": "codelewm.checkpoint.v1"
+  },
+  "migration_hook": null,
+  "schema_version": "codelewm.checkpoint.v1"
+}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/config.json

@@ -0,0 +1,54 @@
+{
+  "data": {
+    "manifest": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/pack/manifest.json",
+    "train": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/pack/hdf5/train.hdf5",
+    "val": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/pack/hdf5/val.hdf5"
+  },
+  "loader": {
+    "batch_size": 64,
+    "num_workers": 4,
+    "persistent_workers": true,
+    "pin_memory": true,
+    "shuffle": true
+  },
+  "loss": {
+    "action_use_margin": 0.02,
+    "action_use_margin_weight": 0.25,
+    "enable_action_use_margin": true,
+    "enable_retrieval_loss": true,
+    "retrieval_temperature": 0.1,
+    "retrieval_weight": 0.05,
+    "sigreg_knots": 17,
+    "sigreg_num_proj": 1024,
+    "sigreg_weight": 0.09
+  },
+  "name": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g",
+  "optimizer": {
+    "lr": 0.0001,
+    "type": "AdamW",
+    "weight_decay": 0.001
+  },
+  "output": {
+    "checkpoint_dir": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/train/checkpoints",
+    "manifest_path": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/train/training_manifest.json",
+    "metrics_path": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/train/metrics.jsonl",
+    "run_dir": "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/train"
+  },
+  "schema_version": "codelewm.train_config.v1",
+  "seed": 240119,
+  "trainer": {
+    "accelerator": "gpu",
+    "devices": 1,
+    "gradient_clip_val": 1.0,
+    "max_steps": 60000,
+    "precision": "bf16-mixed"
+  },
+  "wm": {
+    "action_sequence_length": 256,
+    "action_view": "text",
+    "embed_dim": 256,
+    "history_size": 1,
+    "num_preds": 1,
+    "state_sequence_length": 1024
+  }
+}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/manifest.json

@@ -0,0 +1,115 @@
+{
+  "artifact_id": "training_run-924cd056375f11ea",
+  "artifact_kind": "training_run",
+  "command": [
+    "codelewm",
+    "train",
+    "--config",
+    "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/configs/train.json",
+    "--executor",
+    "torch",
+    "--device",
+    "auto",
+    "--overwrite",
+    "--json",
+    "--log-jsonl",
+    "/CodeLeWM/.artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/logs/train.jsonl"
+  ],
+  "config_sha256": "a8c40e6dad64a96410b9855ce61146a85aa336325980c7a1080943b227e1ffc2",
+  "created_at": "2026-05-20T13:13:45Z",
+  "files": [
+    {
+      "bytes": 1789,
+      "path": "config.json",
+      "sha256": "4e2ad375b3fc132229ada18118a577dc882d68704cb9b317735b03579563f532"
+    },
+    {
+      "bytes": 1472,
+      "path": "metrics.jsonl",
+      "sha256": "9d8b35638953b0648c66d4f69e3e00d554a5c7efb76dd9afa553cce27d6a9eb3"
+    },
+    {
+      "bytes": 1614,
+      "path": "reports/metrics_report.json",
+      "sha256": "74270302bccf8913005b5177d8157266a9ba3f55746e296e2fa99fce66ed70de"
+    },
+    {
+      "bytes": 351495901,
+      "path": "checkpoints/checkpoint.pt",
+      "sha256": "0cb4daf1500495579f5c59cc9fd8aa39f5f70e88f55c0c121320d023b43ddeda"
+    },
+    {
+      "bytes": 511,
+      "path": "checkpoints/checkpoint.pt.manifest.json",
+      "sha256": "9b8cb47facfa9f3c721fa845df0e374aa16692a63f4a93a06b806a95eb4ffeb9"
+    },
+    {
+      "bytes": 2667,
+      "path": "reports/torch_training_report.json",
+      "sha256": "7f2a603f05fd92497c5aaa854d9ba73476c285030176ed068afb62976bb0ec4b"
+    }
+  ],
+  "metadata": {
+    "dataset_manifest_path": ".artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/pack/manifest.json",
+    "executor": {
+      "checkpoint_schema_version": "codelewm.checkpoint.v1",
+      "device": "cuda",
+      "executor": "torch",
+      "objective": {
+        "action_use_margin": 0.02,
+        "action_use_margin_weight": 0.25,
+        "enable_action_use_margin": true,
+        "enable_retrieval_loss": true,
+        "retrieval_temperature": 0.1,
+        "retrieval_weight": 0.05,
+        "sigreg_knots": 17,
+        "sigreg_num_proj": 1024,
+        "sigreg_weight": 0.09
+      },
+      "precision": "bf16-mixed",
+      "torch": "2.12.0+cu130",
+      "train_rows": 18019,
+      "val_rows": 1291
+    },
+    "final_metrics": {
+      "collapse/effective_rank": 10.542142224670206,
+      "collapse/effective_rank_ratio": 0.04118024306511799,
+      "collapse/embedding_count": 192.0,
+      "collapse/embedding_norm_mean": 13.19682605659637,
+      "collapse/latent_dim": 256.0,
+      "collapse/nearest_neighbor_entropy": 4.847709015203599,
+      "collapse/pairwise_cosine_mean": 0.004205391777077934,
+      "collapse/per_dim_variance_max": 2.3262929385455555,
+      "collapse/per_dim_variance_median": 0.6143968449321122,
+      "collapse/per_dim_variance_min": 0.06425978866539729,
+      "loss/action_use_margin": 0.005474980920553207,
+      "loss/action_use_margin_weighted": 0.0013687452301383018,
+      "loss/prediction_mse": 0.006972212344408035,
+      "loss/retrieval": 0.20804595947265625,
+      "loss/retrieval_weighted": 0.010402298532426357,
+      "loss/sigreg": 0.91015625,
+      "loss/sigreg_weighted": 0.08203125,
+      "loss/total": 0.10077450424432755,
+      "train/examples": 3833852.0,
+      "train/examples_per_second": 1064.5055633910936,
+      "train/gradient_norm": 1.3148689270019531,
+      "val/loss/action_use_margin": 0.05166880839637348,
+      "val/loss/action_use_margin_weighted": 0.01291720209909337,
+      "val/loss/prediction_mse": 0.18427681780996777,
+      "val/loss/retrieval": 0.9156766051337832,
+      "val/loss/retrieval_weighted": 0.04578383116140252,
+      "val/loss/sigreg": 1.887503743171692,
+      "val/loss/sigreg_weighted": 0.16987534293106624,
+      "val/loss/total": 0.4128531955537342
+    },
+    "run_id": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g",
+    "schema_version": "codelewm.training_run.v1",
+    "seed": 240119,
+    "step_count": 60000
+  },
+  "parent_artifacts": [
+    "dataset-5695087296ce4a97"
+  ],
+  "schema_version": "codelewm.artifact_manifest.v1",
+  "source_git_sha": "7895d185e165a917af0956a313d8948c04b33638"
+}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/metrics.jsonl

@@ -0,0 +1 @@
+{"metrics": {"collapse/effective_rank": 10.542142224670206, "collapse/effective_rank_ratio": 0.04118024306511799, "collapse/embedding_count": 192.0, "collapse/embedding_norm_mean": 13.19682605659637, "collapse/latent_dim": 256.0, "collapse/nearest_neighbor_entropy": 4.847709015203599, "collapse/pairwise_cosine_mean": 0.004205391777077934, "collapse/per_dim_variance_max": 2.3262929385455555, "collapse/per_dim_variance_median": 0.6143968449321122, "collapse/per_dim_variance_min": 0.06425978866539729, "loss/action_use_margin": 0.005474980920553207, "loss/action_use_margin_weighted": 0.0013687452301383018, "loss/prediction_mse": 0.006972212344408035, "loss/retrieval": 0.20804595947265625, "loss/retrieval_weighted": 0.010402298532426357, "loss/sigreg": 0.91015625, "loss/sigreg_weighted": 0.08203125, "loss/total": 0.10077450424432755, "train/examples": 3833852.0, "train/examples_per_second": 1064.5055633910936, "train/gradient_norm": 1.3148689270019531, "val/loss/action_use_margin": 0.05166880839637348, "val/loss/action_use_margin_weighted": 0.01291720209909337, "val/loss/prediction_mse": 0.18427681780996777, "val/loss/retrieval": 0.9156766051337832, "val/loss/retrieval_weighted": 0.04578383116140252, "val/loss/sigreg": 1.887503743171692, "val/loss/sigreg_weighted": 0.16987534293106624, "val/loss/total": 0.4128531955537342}, "run_id": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g", "schema_version": "codelewm.training_metrics.v1", "step": 60000}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/reports/metrics_report.json

@@ -0,0 +1,36 @@
+{
+  "final_metrics": {
+    "collapse/effective_rank": 10.542142224670206,
+    "collapse/effective_rank_ratio": 0.04118024306511799,
+    "collapse/embedding_count": 192.0,
+    "collapse/embedding_norm_mean": 13.19682605659637,
+    "collapse/latent_dim": 256.0,
+    "collapse/nearest_neighbor_entropy": 4.847709015203599,
+    "collapse/pairwise_cosine_mean": 0.004205391777077934,
+    "collapse/per_dim_variance_max": 2.3262929385455555,
+    "collapse/per_dim_variance_median": 0.6143968449321122,
+    "collapse/per_dim_variance_min": 0.06425978866539729,
+    "loss/action_use_margin": 0.005474980920553207,
+    "loss/action_use_margin_weighted": 0.0013687452301383018,
+    "loss/prediction_mse": 0.006972212344408035,
+    "loss/retrieval": 0.20804595947265625,
+    "loss/retrieval_weighted": 0.010402298532426357,
+    "loss/sigreg": 0.91015625,
+    "loss/sigreg_weighted": 0.08203125,
+    "loss/total": 0.10077450424432755,
+    "train/examples": 3833852.0,
+    "train/examples_per_second": 1064.5055633910936,
+    "train/gradient_norm": 1.3148689270019531,
+    "val/loss/action_use_margin": 0.05166880839637348,
+    "val/loss/action_use_margin_weighted": 0.01291720209909337,
+    "val/loss/prediction_mse": 0.18427681780996777,
+    "val/loss/retrieval": 0.9156766051337832,
+    "val/loss/retrieval_weighted": 0.04578383116140252,
+    "val/loss/sigreg": 1.887503743171692,
+    "val/loss/sigreg_weighted": 0.16987534293106624,
+    "val/loss/total": 0.4128531955537342
+  },
+  "run_id": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g",
+  "schema_version": "codelewm.training_metrics.v1",
+  "step_count": 60000
+}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/reports/torch_training_report.json

@@ -0,0 +1,71 @@
+{
+  "collapse_report": {
+    "effective_rank": 10.542142224670206,
+    "effective_rank_ratio": 0.04118024306511799,
+    "embedding_count": 192,
+    "embedding_norm_mean": 13.19682605659637,
+    "latent_dim": 256,
+    "nearest_neighbor_entropy": 4.847709015203599,
+    "pairwise_cosine_mean": 0.004205391777077934,
+    "per_dim_variance_max": 2.3262929385455555,
+    "per_dim_variance_median": 0.6143968449321122,
+    "per_dim_variance_min": 0.06425978866539729,
+    "schema_version": "codelewm.eval.collapse_report.v1"
+  },
+  "dataset": {
+    "action_view": "text",
+    "train_rows": 18019,
+    "val_rows": 1291
+  },
+  "metrics": {
+    "collapse/effective_rank": 10.542142224670206,
+    "collapse/effective_rank_ratio": 0.04118024306511799,
+    "collapse/embedding_count": 192.0,
+    "collapse/embedding_norm_mean": 13.19682605659637,
+    "collapse/latent_dim": 256.0,
+    "collapse/nearest_neighbor_entropy": 4.847709015203599,
+    "collapse/pairwise_cosine_mean": 0.004205391777077934,
+    "collapse/per_dim_variance_max": 2.3262929385455555,
+    "collapse/per_dim_variance_median": 0.6143968449321122,
+    "collapse/per_dim_variance_min": 0.06425978866539729,
+    "loss/action_use_margin": 0.005474980920553207,
+    "loss/action_use_margin_weighted": 0.0013687452301383018,
+    "loss/prediction_mse": 0.006972212344408035,
+    "loss/retrieval": 0.20804595947265625,
+    "loss/retrieval_weighted": 0.010402298532426357,
+    "loss/sigreg": 0.91015625,
+    "loss/sigreg_weighted": 0.08203125,
+    "loss/total": 0.10077450424432755,
+    "train/examples": 3833852.0,
+    "train/examples_per_second": 1064.5055633910936,
+    "train/gradient_norm": 1.3148689270019531,
+    "val/loss/action_use_margin": 0.05166880839637348,
+    "val/loss/action_use_margin_weighted": 0.01291720209909337,
+    "val/loss/prediction_mse": 0.18427681780996777,
+    "val/loss/retrieval": 0.9156766051337832,
+    "val/loss/retrieval_weighted": 0.04578383116140252,
+    "val/loss/sigreg": 1.887503743171692,
+    "val/loss/sigreg_weighted": 0.16987534293106624,
+    "val/loss/total": 0.4128531955537342
+  },
+  "objective": {
+    "action_use_margin": 0.02,
+    "action_use_margin_weight": 0.25,
+    "enable_action_use_margin": true,
+    "enable_retrieval_loss": true,
+    "retrieval_temperature": 0.1,
+    "retrieval_weight": 0.05,
+    "sigreg_knots": 17,
+    "sigreg_num_proj": 1024,
+    "sigreg_weight": 0.09
+  },
+  "run_id": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g",
+  "runtime": {
+    "device": "cuda",
+    "dtype": "torch.bfloat16",
+    "precision": "bf16-mixed",
+    "torch": "2.12.0+cu130"
+  },
+  "schema_version": "codelewm.torch_training_report.v1",
+  "step_count": 60000
+}

checkpoints/codelewm-action-use-retrieval-20260520-7895d18/training_manifest.json

@@ -0,0 +1,88 @@
+{
+  "artifact_manifest_id": "training_run-924cd056375f11ea",
+  "artifact_manifest_path": "manifest.json",
+  "checkpoint_files": [
+    {
+      "bytes": 351495901,
+      "path": "checkpoints/checkpoint.pt",
+      "sha256": "0cb4daf1500495579f5c59cc9fd8aa39f5f70e88f55c0c121320d023b43ddeda"
+    },
+    {
+      "bytes": 511,
+      "path": "checkpoints/checkpoint.pt.manifest.json",
+      "sha256": "9b8cb47facfa9f3c721fa845df0e374aa16692a63f4a93a06b806a95eb4ffeb9"
+    }
+  ],
+  "config_path": "config.json",
+  "config_sha256": "a8c40e6dad64a96410b9855ce61146a85aa336325980c7a1080943b227e1ffc2",
+  "dataset_manifest_path": ".artifacts/hf/codelewm-action-use-retrieval-20260520-7895d18/pack/manifest.json",
+  "final_metrics": {
+    "collapse/effective_rank": 10.542142224670206,
+    "collapse/effective_rank_ratio": 0.04118024306511799,
+    "collapse/embedding_count": 192.0,
+    "collapse/embedding_norm_mean": 13.19682605659637,
+    "collapse/latent_dim": 256.0,
+    "collapse/nearest_neighbor_entropy": 4.847709015203599,
+    "collapse/pairwise_cosine_mean": 0.004205391777077934,
+    "collapse/per_dim_variance_max": 2.3262929385455555,
+    "collapse/per_dim_variance_median": 0.6143968449321122,
+    "collapse/per_dim_variance_min": 0.06425978866539729,
+    "loss/action_use_margin": 0.005474980920553207,
+    "loss/action_use_margin_weighted": 0.0013687452301383018,
+    "loss/prediction_mse": 0.006972212344408035,
+    "loss/retrieval": 0.20804595947265625,
+    "loss/retrieval_weighted": 0.010402298532426357,
+    "loss/sigreg": 0.91015625,
+    "loss/sigreg_weighted": 0.08203125,
+    "loss/total": 0.10077450424432755,
+    "train/examples": 3833852.0,
+    "train/examples_per_second": 1064.5055633910936,
+    "train/gradient_norm": 1.3148689270019531,
+    "val/loss/action_use_margin": 0.05166880839637348,
+    "val/loss/action_use_margin_weighted": 0.01291720209909337,
+    "val/loss/prediction_mse": 0.18427681780996777,
+    "val/loss/retrieval": 0.9156766051337832,
+    "val/loss/retrieval_weighted": 0.04578383116140252,
+    "val/loss/sigreg": 1.887503743171692,
+    "val/loss/sigreg_weighted": 0.16987534293106624,
+    "val/loss/total": 0.4128531955537342
+  },
+  "metadata": {
+    "executor": {
+      "checkpoint_schema_version": "codelewm.checkpoint.v1",
+      "device": "cuda",
+      "executor": "torch",
+      "objective": {
+        "action_use_margin": 0.02,
+        "action_use_margin_weight": 0.25,
+        "enable_action_use_margin": true,
+        "enable_retrieval_loss": true,
+        "retrieval_temperature": 0.1,
+        "retrieval_weight": 0.05,
+        "sigreg_knots": 17,
+        "sigreg_num_proj": 1024,
+        "sigreg_weight": 0.09
+      },
+      "precision": "bf16-mixed",
+      "torch": "2.12.0+cu130",
+      "train_rows": 18019,
+      "val_rows": 1291
+    }
+  },
+  "metrics_path": "metrics.jsonl",
+  "metrics_report_path": "reports/metrics_report.json",
+  "parent_artifacts": [
+    "dataset-5695087296ce4a97"
+  ],
+  "report_files": [
+    {
+      "bytes": 2667,
+      "path": "reports/torch_training_report.json",
+      "sha256": "7f2a603f05fd92497c5aaa854d9ba73476c285030176ed068afb62976bb0ec4b"
+    }
+  ],
+  "run_id": "codelewm_scaled_action_use_margin_retrieval_gpu_a10g",
+  "schema_version": "codelewm.training_run.v1",
+  "seed": 240119,
+  "step_count": 60000
+}