Add post-training evals, /dev/shm checkpoints, async HF push, and _orig_mod fix

train_all.py:
- Fix torch.compile _orig_mod prefix bug: unwrap model before saving
- Add --run-evals for automatic probes, diagnostics, and Lichess eval
- Add --publish-results to push eval results to HuggingFace
- Add --shm-checkpoints for RAM-backed volatile storage (requires --hf-repo)
- Push checkpoints to HF in background threads (non-blocking)
- Track best val checkpoint for eval and /dev/shm cleanup
- Clean up old /dev/shm checkpoints after successful HF push

eval_probes.py: handle directory-based safetensors checkpoints
monitor_training.sh: new script for remote training monitoring
CLAUDE.md: document --shm-checkpoints, --run-evals, async HF push

CLAUDE.md

@@ -86,7 +86,7 @@ All adapters freeze the backbone and initialize to identity (zero-init or gamma=
 ## Scripts (`scripts/`)
 
 - `train.py` -- Pretrain from scratch (`--variant small|base|large|toy`)
-- `train_all.py` -- Train small/base/large simultaneously on shared data batches
+- `train_all.py` -- Train small/base/large simultaneously on shared data batches. Supports `--run-evals` for automatic post-training probes, diagnostics, and Lichess eval, and `--publish-results` to push eval results to HF.
 - `train_bottleneck.py`, `train_film.py`, `train_lora.py`, `train_sparse.py`, `train_hybrid.py` -- Adapter behavioral cloning on Lichess PGN
 - `train_tiny.py` -- Standalone tiny transformer baseline (no frozen backbone)
 - `eval_accuracy.py` -- MAIA-compatible evaluation (per-phase, per-ply accuracy)
@@ -151,7 +151,9 @@ All training scripts require one of:
 - `--hf-repo REPO_ID` — push checkpoints to a HuggingFace branch as they're written (durable)
 - `--local-checkpoints` — save locally only (for development without an HF account)
 
-HF mode creates a `run/{run_id}` branch. Squash-merge into main when satisfied.
+HF mode creates a `run/{run_id}` branch. HF pushes happen in background threads (one per model slot) so training is not blocked by uploads. Squash-merge into main when satisfied.
+
+Optional: `--shm-checkpoints` writes checkpoints to `/dev/shm` (RAM-backed filesystem, instant writes). Requires `--hf-repo` since `/dev/shm` is volatile. Old checkpoints are cleaned up after successful HF push, keeping only the latest and the best (by val loss) for post-training evals.
 
 ### Data Integrity
 

scripts/eval_probes.py

@@ -61,7 +61,11 @@ def main():
         run_dir = config_path.parent
         if args.run and run_dir.name != args.run:
             continue
-        checkpoints = sorted(run_dir.glob("checkpoints/step_*.pt"))
+        # Find checkpoints: directory-based (safetensors) or legacy .pt
+        checkpoints = sorted(
+            [d for d in run_dir.glob("checkpoints/step_*") if d.is_dir()]
+            or list(run_dir.glob("checkpoints/step_*.pt"))
+        )
         if not checkpoints:
             continue
         latest = checkpoints[-1]

scripts/monitor_training.sh

@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+# Monitor multi-model training: check pod log + HuggingFace checkpoints.
+# Usage: bash scripts/monitor_training.sh <host> <port>
+set -euo pipefail
+
+HOST="${1:-50.145.48.110}"
+PORT="${2:-13321}"
+SSH="ssh -o StrictHostKeyChecking=accept-new -o ConnectTimeout=10 -p $PORT root@$HOST"
+
+echo "=== Training Log ==="
+$SSH "tail -15 /workspace/logs/train_all.log" 2>/dev/null || echo "  (SSH failed)"
+
+echo ""
+echo "=== Process Status ==="
+$SSH "pgrep -f train_all > /dev/null && echo RUNNING || echo STOPPED" 2>/dev/null || echo "  (SSH failed)"
+
+echo ""
+echo "=== HuggingFace Checkpoints ==="
+uv run python3 -c "
+from huggingface_hub import HfApi
+api = HfApi()
+for variant in ['small', 'base', 'large']:
+    repo = f'thomas-schweich/pawn-{variant}'
+    try:
+        branches = [b.name for b in api.list_repo_refs(repo, repo_type='model').branches if b.name.startswith('run/')]
+        for branch in branches:
+            files = [f.rfilename for f in api.list_repo_tree(repo, revision=branch, repo_type='model', recursive=True) if hasattr(f, 'rfilename') and 'checkpoints/' in f.rfilename]
+            ckpts = sorted(set(f.split('/')[1] for f in files if f.startswith('checkpoints/step_')))
+            print(f'  {repo}@{branch}: {len(ckpts)} checkpoints ({ckpts[-1] if ckpts else \"none\"})')
+        if not branches:
+            print(f'  {repo}: no run branches')
+    except Exception as e:
+        print(f'  {repo}: {e}')
+" 2>/dev/null || echo "  (HF check failed)"
+
+echo ""
+echo "=== Metrics Sync ==="
+rsync -az --include='*/' --include='metrics.jsonl' --include='config.json' --exclude='*' \
+    -e "ssh -o StrictHostKeyChecking=accept-new -p $PORT" \
+    "root@$HOST:/workspace/logs/" logs/ 2>/dev/null && echo "  Synced" || echo "  (Sync failed)"

scripts/train_all.py

@@ -15,9 +15,11 @@ import argparse
 import json
 import math
 import os
+import shutil
 import signal
 import sys
 import time
+from pathlib import Path
 
 import torch
 import torch.multiprocessing as mp
@@ -26,7 +28,7 @@ from torch.utils.data import DataLoader
 from pawn.config import CLMConfig, TrainingConfig
 from pawn.model import PAWNCLM, clm_loss
 from pawn.data import CLMDataset, create_validation_set
-from pawn.gpu import configure_gpu, apply_gpu_config
+from pawn.gpu import configure_gpu
 from pawn.checkpoint import save_pretrain_checkpoint, push_checkpoint_to_hf
 
 
@@ -44,12 +46,14 @@ class ModelSlot:
         train_cfg: TrainingConfig,
         device: str,
         hf_repo: str | None,
+        shm_checkpoints: bool = False,
     ):
         self.name = name
         self.model_cfg = model_cfg
         self.train_cfg = train_cfg
         self.device = device
         self.hf_repo = hf_repo
+        self.shm_checkpoints = shm_checkpoints
 
         self.model = PAWNCLM(model_cfg).to(device)
         param_count = sum(p.numel() for p in self.model.parameters())
@@ -70,9 +74,14 @@ class ModelSlot:
 
         self.scaler = torch.amp.GradScaler(device, enabled=train_cfg.use_amp)
 
-        # Run directory
+        # Run directory (logs always on persistent disk)
         self.run_dir = _make_run_dir(train_cfg.log_dir, name)
-        self.checkpoint_dir = os.path.join(self.run_dir, "checkpoints")
+
+        # Checkpoint directory: /dev/shm if requested, else under run_dir
+        if shm_checkpoints:
+            self.checkpoint_dir = f"/dev/shm/pawn_checkpoints/{name}"
+        else:
+            self.checkpoint_dir = os.path.join(self.run_dir, "checkpoints")
         os.makedirs(self.checkpoint_dir, exist_ok=True)
 
         self.jsonl_path = os.path.join(self.run_dir, "metrics.jsonl")
@@ -80,6 +89,13 @@ class ModelSlot:
 
         self.hf_branch = f"run/{os.path.basename(self.run_dir)}" if hf_repo else None
         self.global_step = 0
+        self.best_val_step = 0
+        self.best_val_loss = float("inf")
+
+        # Background HF push (one thread per slot, so pushes don't block training)
+        from concurrent.futures import ThreadPoolExecutor
+        self._hf_push_pool = ThreadPoolExecutor(max_workers=1, thread_name_prefix=f"hf-{name}")
+        self._hf_push_future = None
 
         # Write config
         import subprocess
@@ -133,24 +149,56 @@ class ModelSlot:
         self.scheduler.step()
         return grad_norm
 
+    def _unwrapped_model(self):
+        """Return the unwrapped model (strips torch.compile wrapper)."""
+        m = self.model
+        while hasattr(m, '_orig_mod'):
+            m = m._orig_mod
+        return m
+
     def save_checkpoint(self):
         path = os.path.join(self.checkpoint_dir, f"step_{self.global_step:08d}")
         save_pretrain_checkpoint(
-            path, self.model, self.optimizer, self.scheduler, self.scaler,
+            path, self._unwrapped_model(), self.optimizer, self.scheduler, self.scaler,
             self.global_step, self.model_cfg.__dict__, self.train_cfg.__dict__,
         )
         print(f"  [{self.name}] Checkpoint saved: {path}")
 
         if self.hf_repo and self.hf_branch:
+            self._push_to_hf_async(path, self.global_step)
+
+    def _push_to_hf_async(self, ckpt_path: str, step: int):
+        """Push checkpoint to HuggingFace in a background thread."""
+        # Wait for any previous push to finish before starting a new one
+        if self._hf_push_future is not None:
+            self._hf_push_future.result()  # raises if previous push failed
+
+        def _push():
             try:
                 push_checkpoint_to_hf(
-                    path, self.hf_repo, self.hf_branch,
-                    metrics_path=self.jsonl_path, step=self.global_step,
+                    ckpt_path, self.hf_repo, self.hf_branch,
+                    metrics_path=self.jsonl_path, step=step,
                 )
                 print(f"  [{self.name}] Pushed to HF: {self.hf_repo}@{self.hf_branch}")
+
+                # On /dev/shm, clean up old checkpoints after successful push.
+                # Keep the latest (just saved) and the best (for post-training evals).
+                if self.shm_checkpoints:
+                    keep = {Path(ckpt_path).name, f"step_{self.best_val_step:08d}"}
+                    for old in sorted(Path(self.checkpoint_dir).glob("step_*")):
+                        if old.name not in keep:
+                            shutil.rmtree(old, ignore_errors=True)
             except Exception as e:
                 print(f"  [{self.name}] WARNING: HF push failed: {e}")
 
+        self._hf_push_future = self._hf_push_pool.submit(_push)
+
+    def wait_for_push(self):
+        """Block until any in-flight HF push completes."""
+        if self._hf_push_future is not None:
+            self._hf_push_future.result()
+            self._hf_push_future = None
+
     @torch.no_grad()
     def evaluate(self, val_data: dict[str, torch.Tensor]) -> dict[str, float]:
         self.model.eval()
@@ -182,6 +230,8 @@ class ModelSlot:
         self._jsonl_file.flush()
 
     def close(self):
+        self.wait_for_push()
+        self._hf_push_pool.shutdown(wait=True)
         if self._jsonl_file:
             self._jsonl_file.close()
             self._jsonl_file = None
@@ -215,12 +265,120 @@ def parse_args():
     ckpt_group.add_argument("--hf-repo", type=str, default=None,
                             help="HF repo prefix (appends -{variant}). E.g. thomas-schweich/pawn")
     ckpt_group.add_argument("--local-checkpoints", action="store_true")
+
+    p.add_argument("--shm-checkpoints", action="store_true",
+                    help="Write checkpoints to /dev/shm (RAM-backed, instant writes). "
+                         "Requires --hf-repo since /dev/shm is volatile.")
+
+    p.add_argument("--run-evals", action="store_true",
+                    help="Run probes, diagnostics, and Lichess eval after training completes")
+    p.add_argument("--lichess-pgn", type=str, default=None,
+                    help="Path to Lichess PGN file for eval (required with --run-evals)")
+    p.add_argument("--publish-results", action="store_true",
+                    help="Push eval results to HuggingFace (requires --hf-repo and --run-evals)")
     return p.parse_args()
 
 
+def _run_post_training_evals(slots: list[ModelSlot], args):
+    """Run probes, diagnostics, and Lichess eval on best checkpoint per variant."""
+    import tempfile
+    from pawn.eval_suite.probes import extract_probe_data, train_all_probes
+    from pawn.eval_suite.corpus import generate_corpus, load_corpus
+    from pawn.eval_suite.diagnostics import extract_diagnostic_positions, evaluate_diagnostic_positions
+
+    device = args.device or ("cuda" if torch.cuda.is_available() else "cpu")
+
+    for slot in slots:
+        print(f"\n--- Evaluating {slot.name} ---")
+
+        # Use tracked best val step (kept on /dev/shm if shm_checkpoints)
+        best_step = slot.best_val_step
+        best_loss = slot.best_val_loss
+
+        ckpt_path = os.path.join(slot.checkpoint_dir, f"step_{best_step:08d}")
+        if not os.path.isdir(ckpt_path):
+            # Fall back to latest
+            ckpts = sorted(Path(slot.checkpoint_dir).glob("step_*"))
+            ckpt_path = str(ckpts[-1]) if ckpts else None
+
+        if not ckpt_path:
+            print(f"  No checkpoint found, skipping")
+            continue
+
+        print(f"  Best checkpoint: {ckpt_path} (val_loss={best_loss:.4f})")
+
+        # Load model (unwrapped)
+        from pawn.checkpoint import load_backbone_weights
+        state_dict, _ = load_backbone_weights(ckpt_path)
+        model = PAWNCLM(slot.model_cfg).to(device)
+        model.load_state_dict(state_dict)
+        model.eval()
+
+        results = {}
+
+        # 1. Probes
+        print("  Running probes...")
+        train_data = extract_probe_data(2048, 256, seed=12345)
+        val_data = extract_probe_data(512, 256, seed=54321)
+        probe_results = train_all_probes(
+            model, train_data, val_data, device=device,
+            per_layer=True, n_epochs=20, verbose=True,
+        )
+        results["probes"] = probe_results
+        del train_data, val_data
+
+        # 2. Diagnostics
+        print("  Running diagnostics...")
+        with tempfile.TemporaryDirectory() as tmpdir:
+            corpus_path = generate_corpus(tmpdir, n_games=2048, max_ply=255, seed=99999, batch_size=2048)
+            corpus = load_corpus(corpus_path)
+            positions = extract_diagnostic_positions(corpus, min_per_category=200, max_per_category=1000)
+            diag_results = evaluate_diagnostic_positions(model, positions, corpus, device=device)
+            results["diagnostics"] = diag_results
+
+        # 3. Lichess eval (if PGN provided)
+        if args.lichess_pgn:
+            print("  Running Lichess eval...")
+            from pawn.eval_suite.lichess import prepare_lichess_corpus, evaluate_on_lichess
+            lichess_data = prepare_lichess_corpus(args.lichess_pgn, max_games_per_band=1000)
+            lichess_results = evaluate_on_lichess(model, lichess_data, device=device)
+            results["lichess"] = lichess_results
+
+        # Save results
+        results_path = os.path.join(slot.run_dir, "eval_results.json")
+        with open(results_path, "w") as f:
+            json.dump(results, f, indent=2, default=str)
+        print(f"  Results saved: {results_path}")
+
+        # Publish to HF
+        if args.publish_results and slot.hf_repo and slot.hf_branch:
+            from huggingface_hub import HfApi
+            api = HfApi()
+            try:
+                api.upload_file(
+                    path_or_fileobj=results_path,
+                    path_in_repo="eval_results.json",
+                    repo_id=slot.hf_repo,
+                    repo_type="model",
+                    revision=slot.hf_branch,
+                    commit_message=f"Eval results (best step {best_step})",
+                )
+                print(f"  Published to {slot.hf_repo}@{slot.hf_branch}")
+            except Exception as e:
+                print(f"  WARNING: HF publish failed: {e}")
+
+        del model, state_dict
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+
+
 def main():
     args = parse_args()
 
+    if args.shm_checkpoints and not args.hf_repo:
+        print("ERROR: --shm-checkpoints requires --hf-repo (HF is the only durable store)")
+        sys.exit(1)
+
     device = args.device or ("cuda" if torch.cuda.is_available() else "cpu")
     if device == "cuda":
         gpu_cfg = configure_gpu()
@@ -239,6 +397,8 @@ def main():
     print(f"Device: {device}")
     print(f"Batch size: {args.batch_size}")
     print(f"Total steps: {args.total_steps}")
+    if args.shm_checkpoints:
+        print("Checkpoints: /dev/shm (volatile, HF push is durable store)")
     print()
 
     slots: list[ModelSlot] = []
@@ -256,7 +416,8 @@ def main():
         train_cfg.use_wandb = args.wandb
 
         hf_repo = f"{args.hf_repo}-{name}" if args.hf_repo else None
-        slots.append(ModelSlot(name, model_cfg, train_cfg, device, hf_repo))
+        slots.append(ModelSlot(name, model_cfg, train_cfg, device, hf_repo,
+                               shm_checkpoints=args.shm_checkpoints))
 
     # Shared dataset and validation set
     max_ply = 256
@@ -362,6 +523,11 @@ def main():
                     "timestamp": time.time(),
                     **val_metrics,
                 })
+                # Track best for eval and /dev/shm cleanup
+                vl = val_metrics["val/loss"]
+                if vl < slot.best_val_loss:
+                    slot.best_val_loss = vl
+                    slot.best_val_step = global_step
 
         # Checkpoint
         if global_step % args.checkpoint_interval == 0:
@@ -389,6 +555,13 @@ def main():
     for slot in slots:
         slot.close()
 
+    # Post-training evals
+    if args.run_evals:
+        print("\n" + "=" * 60)
+        print("POST-TRAINING EVALUATION")
+        print("=" * 60)
+        _run_post_training_evals(slots, args)
+
     print("\nAll done.")
 
 

uv.lock

@@ -20,6 +20,15 @@ members = [
     "pawn",
 ]
 
+[[package]]
+name = "annotated-doc"
+version = "0.0.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/57/ba/046ceea27344560984e26a590f90bc7f4a75b06701f653222458922b558c/annotated_doc-0.0.4.tar.gz", hash = "sha256:fbcda96e87e9c92ad167c2e53839e57503ecfda18804ea28102353485033faa4", size = 7288, upload-time = "2025-11-10T22:07:42.062Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/1e/d3/26bf1008eb3d2daa8ef4cacc7f3bfdc11818d111f7e2d0201bc6e3b49d45/annotated_doc-0.0.4-py3-none-any.whl", hash = "sha256:571ac1dc6991c450b25a9c2d84a3705e2ae7a53467b5d111c24fa8baabbed320", size = 5303, upload-time = "2025-11-10T22:07:40.673Z" },
+]
+
 [[package]]
 name = "annotated-types"
 version = "0.7.0"
@@ -499,6 +508,70 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/04/4b/29cac41a4d98d144bf5f6d33995617b185d14b22401f75ca86f384e87ff1/h11-0.16.0-py3-none-any.whl", hash = "sha256:63cf8bbe7522de3bf65932fda1d9c2772064ffb3dae62d55932da54b31cb6c86", size = 37515, upload-time = "2025-04-24T03:35:24.344Z" },
 ]
 
+[[package]]
+name = "hf-xet"
+version = "1.4.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/09/08/23c84a26716382c89151b5b447b4beb19e3345f3a93d3b73009a71a57ad3/hf_xet-1.4.2.tar.gz", hash = "sha256:b7457b6b482d9e0743bd116363239b1fa904a5e65deede350fbc0c4ea67c71ea", size = 672357, upload-time = "2026-03-13T06:58:51.077Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b4/86/b40b83a2ff03ef05c4478d2672b1fc2b9683ff870e2b25f4f3af240f2e7b/hf_xet-1.4.2-cp37-abi3-macosx_10_12_x86_64.whl", hash = "sha256:71f02d6e4cdd07f344f6844845d78518cc7186bd2bc52d37c3b73dc26a3b0bc5", size = 3800339, upload-time = "2026-03-13T06:58:36.245Z" },
+    { url = "https://files.pythonhosted.org/packages/64/2e/af4475c32b4378b0e92a587adb1aa3ec53e3450fd3e5fe0372a874531c00/hf_xet-1.4.2-cp37-abi3-macosx_11_0_arm64.whl", hash = "sha256:e9b38d876e94d4bdcf650778d6ebbaa791dd28de08db9736c43faff06ede1b5a", size = 3559664, upload-time = "2026-03-13T06:58:34.787Z" },
+    { url = "https://files.pythonhosted.org/packages/3c/4c/781267da3188db679e601de18112021a5cb16506fe86b246e22c5401a9c4/hf_xet-1.4.2-cp37-abi3-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:77e8c180b7ef12d8a96739a4e1e558847002afe9ea63b6f6358b2271a8bdda1c", size = 4217422, upload-time = "2026-03-13T06:58:27.472Z" },
+    { url = "https://files.pythonhosted.org/packages/68/47/d6cf4a39ecf6c7705f887a46f6ef5c8455b44ad9eb0d391aa7e8a2ff7fea/hf_xet-1.4.2-cp37-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:c3b3c6a882016b94b6c210957502ff7877802d0dbda8ad142c8595db8b944271", size = 3992847, upload-time = "2026-03-13T06:58:25.989Z" },
+    { url = "https://files.pythonhosted.org/packages/2d/ef/e80815061abff54697239803948abc665c6b1d237102c174f4f7a9a5ffc5/hf_xet-1.4.2-cp37-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:9d9a634cc929cfbaf2e1a50c0e532ae8c78fa98618426769480c58501e8c8ac2", size = 4193843, upload-time = "2026-03-13T06:58:44.59Z" },
+    { url = "https://files.pythonhosted.org/packages/54/75/07f6aa680575d9646c4167db6407c41340cbe2357f5654c4e72a1b01ca14/hf_xet-1.4.2-cp37-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:6b0932eb8b10317ea78b7da6bab172b17be03bbcd7809383d8d5abd6a2233e04", size = 4432751, upload-time = "2026-03-13T06:58:46.533Z" },
+    { url = "https://files.pythonhosted.org/packages/cd/71/193eabd7e7d4b903c4aa983a215509c6114915a5a237525ec562baddb868/hf_xet-1.4.2-cp37-abi3-win_amd64.whl", hash = "sha256:ad185719fb2e8ac26f88c8100562dbf9dbdcc3d9d2add00faa94b5f106aea53f", size = 3671149, upload-time = "2026-03-13T06:58:57.07Z" },
+    { url = "https://files.pythonhosted.org/packages/b4/7e/ccf239da366b37ba7f0b36095450efae4a64980bdc7ec2f51354205fdf39/hf_xet-1.4.2-cp37-abi3-win_arm64.whl", hash = "sha256:32c012286b581f783653e718c1862aea5b9eb140631685bb0c5e7012c8719a87", size = 3533426, upload-time = "2026-03-13T06:58:55.46Z" },
+]
+
+[[package]]
+name = "httpcore"
+version = "1.0.9"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "certifi", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "h11", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/06/94/82699a10bca87a5556c9c59b5963f2d039dbd239f25bc2a63907a05a14cb/httpcore-1.0.9.tar.gz", hash = "sha256:6e34463af53fd2ab5d807f399a9b45ea31c3dfa2276f15a2c3f00afff6e176e8", size = 85484, upload-time = "2025-04-24T22:06:22.219Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7e/f5/f66802a942d491edb555dd61e3a9961140fd64c90bce1eafd741609d334d/httpcore-1.0.9-py3-none-any.whl", hash = "sha256:2d400746a40668fc9dec9810239072b40b4484b640a8c38fd654a024c7a1bf55", size = 78784, upload-time = "2025-04-24T22:06:20.566Z" },
+]
+
+[[package]]
+name = "httpx"
+version = "0.28.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "anyio", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "certifi", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "httpcore", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "idna", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b1/df/48c586a5fe32a0f01324ee087459e112ebb7224f646c0b5023f5e79e9956/httpx-0.28.1.tar.gz", hash = "sha256:75e98c5f16b0f35b567856f597f06ff2270a374470a5c2392242528e3e3e42fc", size = 141406, upload-time = "2024-12-06T15:37:23.222Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/2a/39/e50c7c3a983047577ee07d2a9e53faf5a69493943ec3f6a384bdc792deb2/httpx-0.28.1-py3-none-any.whl", hash = "sha256:d909fcccc110f8c7faf814ca82a9a4d816bc5a6dbfea25d6591d6985b8ba59ad", size = 73517, upload-time = "2024-12-06T15:37:21.509Z" },
+]
+
+[[package]]
+name = "huggingface-hub"
+version = "1.7.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "filelock", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "fsspec", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "hf-xet", marker = "(platform_machine != 'AMD64' and platform_machine != 'aarch64' and platform_machine != 'amd64' and platform_machine != 'arm64' and platform_machine != 'x86_64' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm') or (platform_machine == 'AMD64' and sys_platform == 'linux') or (platform_machine == 'aarch64' and sys_platform == 'linux') or (platform_machine == 'amd64' and sys_platform == 'linux') or (platform_machine == 'arm64' and sys_platform == 'linux') or (platform_machine == 'x86_64' and sys_platform == 'linux') or (sys_platform != 'linux' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "httpx", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "packaging", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "pyyaml", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "tqdm", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "typer", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "typing-extensions", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/19/15/eafc1c57bf0f8afffb243dcd4c0cceb785e956acc17bba4d9bf2ae21fc9c/huggingface_hub-1.7.2.tar.gz", hash = "sha256:7f7e294e9bbb822e025bdb2ada025fa4344d978175a7f78e824d86e35f7ab43b", size = 724684, upload-time = "2026-03-20T10:36:08.767Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/08/de/3ad061a05f74728927ded48c90b73521b9a9328c85d841bdefb30e01fb85/huggingface_hub-1.7.2-py3-none-any.whl", hash = "sha256:288f33a0a17b2a73a1359e2a5fd28d1becb2c121748c6173ab8643fb342c850e", size = 618036, upload-time = "2026-03-20T10:36:06.824Z" },
+]
+
 [[package]]
 name = "humanize"
 version = "4.15.0"
@@ -1322,10 +1395,22 @@ name = "pawn"
 version = "0.1.0"
 source = { editable = "." }
 dependencies = [
+    { name = "anywidget", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "chess-engine", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "huggingface-hub", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "ipykernel", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "matplotlib", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "numpy", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "pandas", version = "2.3.3", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version < '3.11' and sys_platform == 'linux') or (python_full_version >= '3.11' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm') or (sys_platform != 'linux' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "pandas", version = "3.0.1", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version >= '3.11' and sys_platform == 'linux') or (python_full_version < '3.11' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm') or (sys_platform != 'linux' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "plotly", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "polars", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "psutil", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "pyarrow", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "pytest", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "safetensors", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "seaborn", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "solara", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "tqdm", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
     { name = "wandb", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
 ]
@@ -1334,23 +1419,6 @@ dependencies = [
 cu128 = [
     { name = "torch", version = "2.10.0+cu128", source = { registry = "https://download.pytorch.org/whl/cu128" }, marker = "sys_platform == 'linux'" },
 ]
-dashboard = [
-    { name = "anywidget", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "pandas", version = "2.3.3", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version < '3.11' and sys_platform == 'linux') or (python_full_version >= '3.11' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm') or (sys_platform != 'linux' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "pandas", version = "3.0.1", source = { registry = "https://pypi.org/simple" }, marker = "(python_full_version >= '3.11' and sys_platform == 'linux') or (python_full_version < '3.11' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm') or (sys_platform != 'linux' and extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "plotly", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "solara", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-]
-dev = [
-    { name = "ipykernel", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "pytest", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-]
-eval = [
-    { name = "matplotlib", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "polars", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "pyarrow", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-    { name = "seaborn", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
-]
 rocm = [
     { name = "torch", version = "2.10.0+rocm7.1", source = { registry = "https://download.pytorch.org/whl/rocm7.1" }, marker = "sys_platform == 'linux'" },
     { name = "triton-rocm", marker = "sys_platform == 'linux'" },
@@ -1358,27 +1426,28 @@ rocm = [
 
 [package.metadata]
 requires-dist = [
-    { name = "anywidget", marker = "extra == 'dashboard'", specifier = ">=0.9.21" },
+    { name = "anywidget", specifier = ">=0.9.21" },
     { name = "chess-engine", editable = "engine" },
-    { name = "ipykernel", marker = "extra == 'dev'", specifier = ">=7.2.0" },
-    { name = "matplotlib", marker = "extra == 'eval'", specifier = ">=3.10.8" },
+    { name = "huggingface-hub", specifier = ">=0.20.0" },
+    { name = "ipykernel", specifier = ">=7.2.0" },
+    { name = "matplotlib", specifier = ">=3.10.8" },
     { name = "numpy", specifier = "~=2.2.0" },
-    { name = "pandas", marker = "extra == 'dashboard'", specifier = ">=2.0.0" },
-    { name = "plotly", marker = "extra == 'dashboard'", specifier = ">=5.18.0" },
-    { name = "polars", marker = "extra == 'eval'", specifier = ">=1.39.0" },
+    { name = "pandas", specifier = ">=2.0.0" },
+    { name = "plotly", specifier = ">=5.18.0" },
+    { name = "polars", specifier = ">=1.39.0" },
     { name = "psutil", specifier = ">=5.9.0" },
-    { name = "pyarrow", marker = "extra == 'eval'", specifier = ">=23.0.1" },
-    { name = "pytest", marker = "extra == 'dev'", specifier = "~=9.0.0" },
+    { name = "pyarrow", specifier = ">=23.0.1" },
+    { name = "pytest", specifier = "~=9.0.0" },
     { name = "safetensors", specifier = ">=0.4.0" },
-    { name = "seaborn", marker = "extra == 'eval'", specifier = ">=0.13.2" },
-    { name = "solara", marker = "extra == 'dashboard'", specifier = ">=1.0.0" },
+    { name = "seaborn", specifier = ">=0.13.2" },
+    { name = "solara", specifier = ">=1.0.0" },
     { name = "torch", marker = "extra == 'cu128'", specifier = "~=2.10.0", index = "https://download.pytorch.org/whl/cu128", conflict = { package = "pawn", extra = "cu128" } },
     { name = "torch", marker = "extra == 'rocm'", specifier = "~=2.10.0", index = "https://download.pytorch.org/whl/rocm7.1", conflict = { package = "pawn", extra = "rocm" } },
     { name = "tqdm", specifier = "~=4.67.0" },
     { name = "triton-rocm", marker = "extra == 'rocm'", specifier = ">=3.6.0", index = "https://download.pytorch.org/whl/rocm7.1", conflict = { package = "pawn", extra = "rocm" } },
     { name = "wandb", specifier = "~=0.25.0" },
 ]
-provides-extras = ["rocm", "cu128", "eval", "dashboard", "dev"]
+provides-extras = ["rocm", "cu128"]
 
 [[package]]
 name = "pexpect"
@@ -2065,6 +2134,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/9d/76/f789f7a86709c6b087c5a2f52f911838cad707cc613162401badc665acfe/setuptools-82.0.1-py3-none-any.whl", hash = "sha256:a59e362652f08dcd477c78bb6e7bd9d80a7995bc73ce773050228a348ce2e5bb", size = 1006223, upload-time = "2026-03-09T12:47:15.026Z" },
 ]
 
+[[package]]
+name = "shellingham"
+version = "1.5.4"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/58/15/8b3609fd3830ef7b27b655beb4b4e9c62313a4e8da8c676e142cc210d58e/shellingham-1.5.4.tar.gz", hash = "sha256:8dbca0739d487e5bd35ab3ca4b36e11c4078f3a234bfce294b0a0291363404de", size = 10310, upload-time = "2023-10-24T04:13:40.426Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/e0/f9/0595336914c5619e5f28a1fb793285925a8cd4b432c9da0a987836c7f822/shellingham-1.5.4-py2.py3-none-any.whl", hash = "sha256:7ecfff8f2fd72616f7481040475a65b2bf8af90a56c89140852d1120324e8686", size = 9755, upload-time = "2023-10-24T04:13:38.866Z" },
+]
+
 [[package]]
 name = "six"
 version = "1.17.0"
@@ -2351,6 +2429,21 @@ wheels = [
     { url = "https://download.pytorch.org/whl/triton_rocm-3.6.0-cp312-cp312-linux_x86_64.whl" },
 ]
 
+[[package]]
+name = "typer"
+version = "0.24.1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "annotated-doc", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "click", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "rich", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+    { name = "shellingham", marker = "sys_platform == 'linux' or (extra == 'extra-4-pawn-cu128' and extra == 'extra-4-pawn-rocm')" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/f5/24/cb09efec5cc954f7f9b930bf8279447d24618bb6758d4f6adf2574c41780/typer-0.24.1.tar.gz", hash = "sha256:e39b4732d65fbdcde189ae76cf7cd48aeae72919dea1fdfc16593be016256b45", size = 118613, upload-time = "2026-02-21T16:54:40.609Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/4a/91/48db081e7a63bb37284f9fbcefda7c44c277b18b0e13fbc36ea2335b71e6/typer-0.24.1-py3-none-any.whl", hash = "sha256:112c1f0ce578bfb4cab9ffdabc68f031416ebcc216536611ba21f04e9aa84c9e", size = 56085, upload-time = "2026-02-21T16:54:41.616Z" },
+]
+
 [[package]]
 name = "typing-extensions"
 version = "4.15.0"