Publish exact DistillationProject training split and provenance

README.md

@@ -0,0 +1,33 @@
+---
+language:
+- en
+task_categories:
+- text-generation
+tags:
+- math
+- sft
+- distillation-project
+source_datasets:
+- AI-MO/NuminaMath-1.5
+configs:
+- config_name: prepared_50k
+  default: true
+  data_files:
+  - split: train
+    path: data/prepared-50000.parquet
+- config_name: retained_45k_checkpoint
+  data_files:
+  - split: train
+    path: data/committed-45000.parquet
+  - split: not_in_checkpoint
+    path: data/not-in-45k-checkpoint-5000.parquet
+license: apache-2.0
+---
+
+# DistillationProject: original NuminaMath 50K pool
+
+The exact 50,000-example prepared pool for the original Qwen2.5-1.5B base SFT experiment. The `prepared_50k` configuration preserves the original prepared parquet byte for byte. Training stopped with a retained 45K checkpoint. The `retained_45k_checkpoint` configuration exposes its 45,000 contributing examples in the actual seed-42 loader order, plus the 5,000 examples not represented in that checkpoint. Intermediate Numina checkpoints correspond to 5K prefixes of the ordered 45K view. The remaining 5K view is a provenance split, not a validation benchmark.
+
+Exact original question and response text is preserved. No resampling, solution rewriting or truncation is performed for publication. Source revisions, row counts, order and file SHA256 values are recorded in `publication.json` and `provenance/`. Prepared tables retain their original columns. Source validity labels do not establish independently verified mathematical correctness.
+
+Collection: https://huggingface.co/collections/zbeeb/disstillation-project-6ace05e5f40dbe715407516c

data/committed-45000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d189768b7dd1e1ec9ea1d29f3523a0dbf8c5d32f266f78bb2295da8b0bca8287
+size 22042171

data/not-in-45k-checkpoint-5000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36ad149ee135e52547315a4ddee98ed47a4e58bf0ad734cf562ea35c32978fac
+size 2490539

data/prepared-50000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:117a4a06ac35eb24bb6a0fa843ef6955d87a187eac38ef97e648da338ab83f2d
+size 24521313

provenance/original/manifest.json

@@ -0,0 +1,74 @@
+{
+  "at": "2026-10-13T00:24:13.659856+00:00",
+  "experiment_sha256": "7cb618ffedcee2ceb662a5080deea283c6596d17be69cce84b11f3e4155c0bd5",
+  "files_sha256": {
+    "data/sft-50000.parquet": "117a4a06ac35eb24bb6a0fa843ef6955d87a187eac38ef97e648da338ab83f2d",
+    "data/train.arrow": "c2221bbde6ae2fc263e72c15dd5255c51e07e353499345f24a5eccb2d7082a8f",
+    "models/base/merges.txt": "599bab54075088774b1733fde865d5bd747cbcc7a547c5bc12610e874e26f5e3",
+    "models/base/vocab.json": "ca10d7e9fb3ed18575dd1e277a2579c16d108e32f27439684afa0e10b1440910",
+    "models/base/config.json": "0e8c8aa86468aba09c9d32157ff4bc2301c7e6c50e4398960425b2ea71e66f77",
+    "models/base/model.safetensors": "a961db72e75d52b18e6b0c9d379e51a26973b233385e0e127fdda7d648aec796",
+    "models/base/tokenizer_config.json": "c91efca15ceff6e9ee9424db58a6f59cd41294e550a86cbd07e3c1fb500b34f9",
+    "models/base/LICENSE": "832dd9e00a68dd83b3c3fb9f5588dad7dcf337a0db50f7d9483f310cd292e92e",
+    "models/base/tokenizer.json": "c0382117ea329cdf097041132f6d735924b697924d6f6fc3945713e96ce87539",
+    "models/base/README.md": "84cc3bdbfd781149913126b70981cf1e341176fe4bc9d878601a83f18b035ddf",
+    "models/base/generation_config.json": "8c970692323e3ea0e9b8b0a4dca79388d31226e41f83c9fd6014804280ebf6e8",
+    "data/raw/numina/data/train-00000-of-00003.parquet": "1e37cedcc5104c0d5a9afbf8b26328a59d7005cd37c2bd08352419f5f0e03214",
+    "data/raw/numina/data/train-00001-of-00003.parquet": "bc48fdddc1ab4e559727ad4f1f03b0263785c1c16b7964fb885fba07e18d6a39",
+    "data/raw/numina/data/train-00002-of-00003.parquet": "951aa899fe2a1b64f5b40c6636372d35b9db4f0b9c922e874d4bf357db64da66",
+    "data/benchmark-exclusions/MATH-500/test.jsonl": "35dc41080a3680858b27fa7e0533d2d547825316fc5dafe5d316f4ccc5a06132",
+    "data/benchmark-exclusions/aime_2024/data/train-00000-of-00001.parquet": "26139847601a5037c237d5928b195e7260ca8074cf4f264b794af42847f79ccf",
+    "data/benchmark-exclusions/amc23/test-00000-of-00001.parquet": "b696e87ba47be4e879a60fd4ef1d4aa522ba78c5ae013c6aff5cc9788a397c5e"
+  },
+  "raw_dataset_sha256": {
+    "data/raw/numina/data/train-00000-of-00003.parquet": "1e37cedcc5104c0d5a9afbf8b26328a59d7005cd37c2bd08352419f5f0e03214",
+    "data/raw/numina/data/train-00001-of-00003.parquet": "bc48fdddc1ab4e559727ad4f1f03b0263785c1c16b7964fb885fba07e18d6a39",
+    "data/raw/numina/data/train-00002-of-00003.parquet": "951aa899fe2a1b64f5b40c6636372d35b9db4f0b9c922e874d4bf357db64da66"
+  },
+  "selected_prompt_ids_sha256": "fe2a01533bee1dc88e00c27c003f929fe1ffa476d756f782ed19712ad401307e",
+  "selection_method": "seeded hash priority over full eligible deduplicated source; filter complete lengths in priority order",
+  "examples": 50000,
+  "optimizer_steps": 790,
+  "snapshots": [
+    {
+      "optimizer_step": 79,
+      "examples_processed": 5000
+    },
+    {
+      "optimizer_step": 158,
+      "examples_processed": 10000
+    },
+    {
+      "optimizer_step": 237,
+      "examples_processed": 15000
+    },
+    {
+      "optimizer_step": 316,
+      "examples_processed": 20000
+    },
+    {
+      "optimizer_step": 395,
+      "examples_processed": 25000
+    },
+    {
+      "optimizer_step": 474,
+      "examples_processed": 30000
+    },
+    {
+      "optimizer_step": 553,
+      "examples_processed": 35000
+    },
+    {
+      "optimizer_step": 632,
+      "examples_processed": 40000
+    },
+    {
+      "optimizer_step": 711,
+      "examples_processed": 45000
+    },
+    {
+      "optimizer_step": 790,
+      "examples_processed": 50000
+    }
+  ]
+}

provenance/original/selection_config.json

@@ -0,0 +1,102 @@
+{
+  "model": {
+    "id": "Qwen/Qwen2.5-1.5B",
+    "revision": "8faed761d45a263340a0528343f099c05c9a4323"
+  },
+  "dataset": {
+    "id": "AI-MO/NuminaMath-1.5",
+    "revision": "1b05109f9e5c1ad06c0663519502416c30b300f8",
+    "split": "train",
+    "available_rows": 896215,
+    "files": [
+      "data/train-00000-of-00003.parquet",
+      "data/train-00001-of-00003.parquet",
+      "data/train-00002-of-00003.parquet"
+    ]
+  },
+  "selection": {
+    "examples": 50000,
+    "seed": 42,
+    "candidate_multiplier": 3,
+    "require_problem_valid": "Yes",
+    "require_solution_valid": "Yes",
+    "exclude_sources": [
+      "amc_aime",
+      "metamath"
+    ],
+    "maximum_prompt_tokens": 1024,
+    "overlength_policy": "reject",
+    "preserve_raw_text": true
+  },
+  "training": {
+    "epochs": 1,
+    "learning_rate": 1e-05,
+    "global_batch_size": 64,
+    "micro_batch_size": 1,
+    "maximum_sequence_tokens": 4096,
+    "warmup_ratio": 0.03,
+    "scheduler": "cosine",
+    "weight_decay": 0.01,
+    "gradient_clip": 1.0,
+    "packing": false,
+    "snapshot_examples": 5000,
+    "keep_all_checkpoints": true,
+    "system_prompt": "Solve the mathematical problem and explain your reasoning.",
+    "loss": "completion_only_token_mean",
+    "finetuning": "full_parameters"
+  },
+  "topology": {
+    "nodes": 2,
+    "gpus_per_node": 8,
+    "minimum_gpu_memory_gib": 79,
+    "maximum_hp_gpus": 18,
+    "master_port": 29571,
+    "timeout_seconds": 1800,
+    "node_gpu_counts": [
+      8,
+      9
+    ],
+    "minimum_gpu_memory_gib_by_node": [
+      79,
+      39
+    ],
+    "secondary_nodes": [
+      "deep-chungus-3"
+    ]
+  },
+  "runtime": {
+    "profile": "prime-rl",
+    "source_hashes": {
+      "prime_rl.trainer.sft.train": "ac5f90594634bf3c57d1770ba6a169ae45fc9f541e25222a27c2e1b0c818e298",
+      "prime_rl.trainer.ckpt": "c7488f3ded3b56bb0633b75b361054d9df135b723bf4d9dd0909672a39a9d7ee"
+    },
+    "expected_packages": {
+      "torch": "2.11.0+cu128",
+      "transformers": "5.6.2"
+    }
+  },
+  "benchmark_exclusions": [
+    {
+      "id": "HuggingFaceH4/MATH-500",
+      "revision": "6e4ed1a2a79af7d8630a6b768ec859cb5af4d3be",
+      "file": "test.jsonl",
+      "question_key": "problem",
+      "count": 500
+    },
+    {
+      "id": "HuggingFaceH4/aime_2024",
+      "revision": "2fe88a2f1091d5048c0f36abc874fb997b3dd99a",
+      "file": "data/train-00000-of-00001.parquet",
+      "question_key": "problem",
+      "count": 30
+    },
+    {
+      "id": "math-ai/amc23",
+      "revision": "80815d37005feb82cd7f8fbc6901d5d3eff43057",
+      "file": "test-00000-of-00001.parquet",
+      "question_key": "question",
+      "count": 40
+    }
+  ],
+  "launch_status": "not_approved"
+}

provenance/original/training_loader.py

@@ -0,0 +1,90 @@
+import os
+import random
+from collections import defaultdict
+
+from core import ROOT, Plan, completion_mask, config, digest, world_size
+
+
+class PreparedLoader:
+    def __init__(self):
+        import pyarrow as pa
+
+        self.cfg = config()
+        self.rank = int(os.environ["RANK"])
+        self.world = int(os.environ["WORLD_SIZE"])
+        expected = world_size(self.cfg)
+        if self.world != expected:
+            raise ValueError("Actual distributed world differs from the prepared topology")
+        training = self.cfg["training"]
+        self.plan = Plan(self.cfg["selection"]["examples"], training["snapshot_examples"], training["global_batch_size"], self.world)
+        self.mapping = pa.memory_map(str(ROOT / "data/train.arrow"), "r")
+        self.table = pa.ipc.open_file(self.mapping).read_all()
+        if len(self.table) != self.plan.examples:
+            raise ValueError("Prepared example count differs")
+        self.order = list(range(len(self.table)))
+        random.Random(self.cfg["selection"]["seed"]).shuffle(self.order)
+        self.cursor = 0
+        self.num_samples, self.num_tokens = defaultdict(int), defaultdict(int)
+        self.input_tokens = 0
+        self.dataset = self
+        self.manifest_hash = digest(ROOT / "data/manifest.json")
+
+    @property
+    def step(self):
+        return self.plan.processed(self.cursor)
+
+    @property
+    def epoch(self):
+        return self.step / self.plan.examples
+
+    def __iter__(self):
+        return self
+
+    def __next__(self):
+        import torch
+
+        if self.cursor == self.plan.total_steps * self.plan.accumulation:
+            raise StopIteration
+        index = self.plan.index(self.cursor, self.rank)
+        dummy = index is None
+        row = self.table.slice(self.order[0 if dummy else index], 1).to_pylist()[0]
+        tokens, prefix = row["input_ids"], row["prompt_length"]
+        if not 0 < prefix < len(tokens) <= self.cfg["training"]["maximum_sequence_tokens"]:
+            raise ValueError("Prepared completion boundary or sequence length is invalid")
+        length = len(tokens) - 1
+        mask = torch.tensor(completion_mask(tokens, prefix, dummy), dtype=torch.bool)
+        if not dummy:
+            self.num_samples["numina"] += 1
+            self.num_tokens["numina"] += int(mask.sum())
+            self.input_tokens += length
+        self.cursor += 1
+        return {"input_ids": torch.tensor(tokens[:-1], dtype=torch.long).unsqueeze(0),
+                "target_ids": torch.tensor(tokens[1:], dtype=torch.long).unsqueeze(0),
+                "position_ids": torch.arange(length, dtype=torch.long).unsqueeze(0),
+                "loss_mask": mask.unsqueeze(0), "seq_lens": torch.tensor([length], dtype=torch.long)}
+
+    def state_dict(self):
+        import numpy as np
+        import torch
+
+        return {"cursor": self.cursor, "rank": self.rank, "world": self.world, "manifest_sha256": self.manifest_hash,
+                "cpu_rng": torch.get_rng_state(), "cuda_rng": torch.cuda.get_rng_state(),
+                "python_rng": random.getstate(), "numpy_rng": np.random.get_state(),
+                "num_samples": dict(self.num_samples), "num_tokens": dict(self.num_tokens), "input_tokens": self.input_tokens}
+
+    def load_state_dict(self, state):
+        import numpy as np
+        import torch
+
+        if (state["rank"], state["world"], state["manifest_sha256"]) != (self.rank, self.world, self.manifest_hash):
+            raise ValueError("Resume rank, topology or dataset contract differs; no fallback to rank zero")
+        if not 0 <= state["cursor"] <= self.plan.total_steps * self.plan.accumulation or state["cursor"] % self.plan.accumulation:
+            raise ValueError("Saved cursor is not at an optimizer boundary")
+        self.cursor = state["cursor"]
+        self.num_samples.update(state["num_samples"])
+        self.num_tokens.update(state["num_tokens"])
+        self.input_tokens = state["input_tokens"]
+        torch.set_rng_state(state["cpu_rng"])
+        torch.cuda.set_rng_state(state["cuda_rng"])
+        random.setstate(state["python_rng"])
+        np.random.set_state(state["numpy_rng"])

provenance/retained_45k_export.json

@@ -0,0 +1,28 @@
+{
+  "at": "2026-10-13T22:26:35.779656+00:00",
+  "teacher_dataset": {
+    "id": "AI-MO/NuminaMath-1.5",
+    "revision": "1b05109f9e5c1ad06c0663519502416c30b300f8",
+    "split": "train",
+    "available_rows": 896215,
+    "files": [
+      "data/train-00000-of-00003.parquet",
+      "data/train-00001-of-00003.parquet",
+      "data/train-00002-of-00003.parquet"
+    ]
+  },
+  "initial_model": {
+    "id": "Qwen/Qwen2.5-1.5B",
+    "revision": "8faed761d45a263340a0528343f099c05c9a4323"
+  },
+  "optimizer_step": 711,
+  "examples_processed": 45000,
+  "contract_id": "eeb34a30b8666f32f58755d2ad3f1d969f79cbd0d90ed7d4ecbf13b100d0ac14",
+  "native_metadata_sha256": "2dfac97a4f7621236340491cc0946e7c62aaeead005f4cb0387265891dbc02b1",
+  "weights_sha256": {
+    "model-00001-of-00002.safetensors": "8de13194076e91ba55881c618fa89913e0963dff2b21be29398737705f6b2c52",
+    "model-00002-of-00002.safetensors": "51894ebf01457a02fae8a73b9a2d6c8e43d4644a77333f86564a8a14ee7196d0"
+  },
+  "saved_precision": "float32",
+  "retention": "all"
+}

publication.json

@@ -0,0 +1,40 @@
+{
+  "publication_version": 1,
+  "raw_question_response_text": "unchanged",
+  "seed": 42,
+  "sources": {
+    "original": {
+      "root": "/mnt/nfs/home/mohamadzbib/projects/DistillationProject",
+      "config": "configs/experiment.json",
+      "config_sha256": "7cb618ffedcee2ceb662a5080deea283c6596d17be69cce84b11f3e4155c0bd5",
+      "manifest_sha256": "616b49cfb824a4099a5ce9a29e0d6bd665cd876d88f31d28a17cd865b5df6edb",
+      "loader_sha256": "adbf05e429945472460b5b69a5652939ac9511b8b630b84a19ff4f73f667ffd2"
+    }
+  },
+  "details": {
+    "prepared_examples": 50000,
+    "retained_checkpoint_examples": 45000,
+    "not_in_retained_45k_checkpoint": 5000,
+    "retained_checkpoint_optimizer_step": 711,
+    "training_order": "seed42 permutation of prepared row indices",
+    "maximum_sequence_tokens": 4096,
+    "overlength_policy": "reject"
+  },
+  "files": {
+    "data/prepared-50000.parquet": {
+      "sha256": "117a4a06ac35eb24bb6a0fa843ef6955d87a187eac38ef97e648da338ab83f2d",
+      "bytes": 24521313,
+      "rows": 50000
+    },
+    "data/committed-45000.parquet": {
+      "sha256": "d189768b7dd1e1ec9ea1d29f3523a0dbf8c5d32f266f78bb2295da8b0bca8287",
+      "bytes": 22042171,
+      "rows": 45000
+    },
+    "data/not-in-45k-checkpoint-5000.parquet": {
+      "sha256": "36ad149ee135e52547315a4ddee98ed47a4e58bf0ad734cf562ea35c32978fac",
+      "bytes": 2490539,
+      "rows": 5000
+    }
+  }
+}