Resumable Kaggle evaluations, sequential SliceGPT LoRA and exported Modal progress

bundle_manifest.json

@@ -1,6 +1,6 @@
 {
   "run_id": "run-9fc515e82f313fab",
-  "zip_sha256": "6097ccf56cdc81066d600c7ab7381b210a5b9fddeb553216d3a0e142819bb3f4",
+  "zip_sha256": "42745741277f2faa095934a94f10580fde89b18b6f59c11dcc8e809f60c65972",
   "available_checkpoints": 27,
   "shared_config": "Pruning-LLMs/diffusion_eval_and_vis/config.py",
   "smoke_tests": "All three methods on tiny LLaDA/Dream: likelihood, generation, CPU/disk and CUDA fp16 offload, reload agreement; task sharding and merging"

finetune_runner.py

@@ -49,11 +49,13 @@ def main():
     p.add_argument("--upload", action="store_true", help="Save/reload/verify, then upload publicly with HF_TOKEN")
     p.add_argument("--namespace", default="vanshnawander")
     args = p.parse_args()
+    from diffusion_eval_and_vis.config import SLICE_TRAINING_EXECUTION
+    execution = dict(SLICE_TRAINING_EXECUTION)
     cfg = StructuredPruningConfig()
     cfg.eval.dtype, cfg.eval.device = args.dtype, args.device
     if args.show_config:
         print(json.dumps({"finetune_config": asdict(cfg.finetune), "eval_config": asdict(cfg.eval),
-                          "dataset_revisions": DATASET_REVISIONS,
+                          "dataset_revisions": DATASET_REVISIONS, "execution": execution,
                           "effective_dataset_order": cfg.finetune.datasets_for_model(args.repo)}, indent=2))
         return
     import torch
@@ -91,7 +93,7 @@ def main():
         print(json.dumps(uploaded, indent=2))
         return
     result = training_core.train_one(asdict(cfg), identifier, job, upload, DATASET_REVISIONS,
-                                    lambda: None, deadline=time.monotonic()+args.time_budget_seconds)
+                                    lambda: None, deadline=time.monotonic()+args.time_budget_seconds, execution=execution)
     print(json.dumps(result, indent=2), flush=True)
     if args.upload and not result.get("continued"):
         if not os.environ.get("HF_TOKEN"):

kaggle_slicegpt_lora.ipynb

@@ -19,7 +19,9 @@
         "# Sequential SliceGPT LoRA on Kaggle\n",
         "This is the same masked-diffusion LoRA procedure as Modal training: one adapter and pinned training datasets. Base models use WikiText-2 \u2192 C4 (2048 optimizer steps). Instruct models use WikiText-2 \u2192 C4 \u2192 Alpaca (3072 steps). All use rank 8/alpha 32, 4096 examples per stage, 2048 tokens and accumulation 4. The setup cell prints the effective dataset order and full shared configuration.\n",
         "\n",
-        "Enable Internet and GPU. On T4 choose both GPUs; the trainer requires all weights on GPUs, with no CPU/disk training offload. FP16 is an explicit precision change; BF16 is available on supporting GPUs. Hyperparameters stay unchanged. Every 50 steps and stage boundary saves optimizer, adapter and RNG state. A session time budget saves and exits so another session can continue.\n"
+        "Enable Internet and GPU. On T4 choose both GPUs; the trainer requires all weights on GPUs, with no CPU/disk training offload. FP16 is an explicit precision change; BF16 is available on supporting GPUs. Hyperparameters stay unchanged. Every 50 steps and stage boundary saves optimizer, adapter and RNG state. A session time budget saves and exits so another session can continue.\n",
+        "\n",
+        "Execution now preloads the tokenized stage datasets onto GPU, tests physical microbatches 4/2/1, and keeps the fastest fit. The effective optimizer batch remains four. CUDA BF16 execution uses autocast and fused AdamW. Actual microbatch, timing and GPU memory are logged; the config cell prints execution settings. Changing GPU/microbatch can change floating-point results and random-number grouping, while sample counts, learning-rate schedule and optimizer cursor are preserved.\n"
       ],
       "id": "cell-0"
     },

portable_eval.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6097ccf56cdc81066d600c7ab7381b210a5b9fddeb553216d3a0e142819bb3f4
-size 67178
+oid sha256:42745741277f2faa095934a94f10580fde89b18b6f59c11dcc8e809f60c65972
+size 71482