Datasets:

Efe2898
/

gemma3-1b-reasoning-stage1-core-4k-tokenized

input_ids listlengths 4.1k 4.1k	length uint16 4.1k 4.1k
[ 2, 105, 2364, 107, 24501, 506, 3755, 529, 614, 40442, 7501, 236787, 41654, 1420, 236782, 236770, 1170, 236800, 1263, 570, 236862, 236778, 236764, 236743, 236812, 236762, 236764, 621, 2989, 2104, 61973, 573, 609, 236762, 2104, 106, 107, 105, ...	4,096
[ 518, 236764, 505, 160670, 577, 4414, 1759, 4945, 236761, 34241, 600, 108, 9873, 107, 236785, 1420, 236782, 236770, 1170, 236746, 236769, 236770, 236862, 236763, 77283, 1420, 236782, 236770, 1170, 236763, 236769, 236770, 236862, 236755, 77283, 142...	4,096
[ 236779, 236800, 3378, 3485, 236884, 236800, 5721, 107, 140560, 107, 67999, 33556, 570, 236779, 236770, 160670, 532, 33556, 570, 236779, 236778, 621, 1473, 107, 44573, 107, 570, 236779, 236770, 570, 236779, 236778, 578, 621, 989, 34432, 1420, ...	4,096
[ 578, 621, 1420, 236782, 236770, 236771, 236771, 1170, 236812, 236771, 236771, 236783, 160670, 108, 29673, 1123, 578, 621, 1420, 236782, 236770, 1170, 236812, 236783, 160670, 108, 29673, 1123, 578, 236743, 236771, 236761, 236778, 236810, 160670, 1...	4,096
[ 236800, 753, 1123, 236884, 236778, 236795, 657, 1156, 3298, 236764, 6590, 600, 1056, 609, 236771, 655, 496, 655, 236743, 236770, 2500, 506, 4901, 609, 236781, 236884, 236778, 900, 496, 236884, 236781, 753, 236743, 236800, 578, 236743, 236771, ...	4,096
[ 518, 681, 236749, 236862, 236770, 236783, 833, 833, 496, 681, 236749, 236862, 236800, 236783, 1704, 107, 833, 833, 518, 681, 236749, 236862, 236778, 236783, 833, 107, 621, 643, 236782, 2513, 236783, 107, 140560, 107, 236743, 107, 236832, 2367...	4,096
[ 692, 1586, 874, 3485, 1631, 3840, 236800, 2500, 236743, 107, 3292, 47157, 621, 54150, 236782, 236755, 39651, 236800, 236764, 236825, 1370, 653, 79094, 236800, 10442, 236825, 54536, 236743, 107, 236769, 236778, 236768, 5612, 44480, 54150, 236782, ...	4,096
[ 236743, 236778, 236761, 236810, 236764, 236743, 236778, 236761, 236810, 236764, 236743, 236800, 236764, 236743, 236800, 236764, 236743, 236800, 236761, 236810, 236764, 236743, 236800, 236761, 236810, 236764, 236743, 236812, 236764, 236743, 236812, 236764...	4,096
[965,568,236781,236884,236812,769,3792,236789,236751,1171,30091,506,5619,573,496,236787,108,236746,5(...TRUNCATED)	4,096
[236760,236769,236762,1223,107,2288,108,236770,236761,5213,66269,87294,569,33556,517,621,62902,23678(...TRUNCATED)	4,096

End of preview. Expand in Data Studio

stage1_core_reasoning_4k

Tokenized packed SFT dataset for google/gemma-3-1b-pt.

Format

Split: train
Columns: input_ids, length
input_ids: packed 4096-token uint32 sequences
Labels are not stored; use labels = input_ids.copy() in the causal LM collator.
Assistant text before tokenization is normalized as:

<think>
reasoning
</think>
final answer

No extra special tokens were added to the tokenizer.

Stats

{
  "stage": "stage1_core_reasoning_4k",
  "tokenizer": "google/gemma-3-1b-pt",
  "seq_len": 4096,
  "max_example_tokens": 16384,
  "target_tokens": 500000000,
  "written_tokens": 500002816,
  "resumed_from_tokens": 0,
  "num_sequences": 122071,
  "num_shards": 15,
  "shard_rows": 8192,
  "parquet_zstd_level": 1,
  "repo_id": "Efe2898/gemma3-1b-reasoning-stage1-core-4k-tokenized",
  "streaming": true,
  "cpu_cores": 4,
  "tokenize_batch_size": 2048,
  "source_batch_size": 2048,
  "weights": {
    "NuminaMath-CoT": 22,
    "OpenR1-Math-220k": 18,
    "GSM8K": 4,
    "NuminaMath-TIR": 8,
    "OpenThoughts-114k": 12,
    "Bespoke-Stratos-17k": 4,
    "turkish-math-186k": 18,
    "gsm8k_tr": 4,
    "OpenThoughts-TR-18k": 7,
    "turkish-general-reasoning-28k": 3
  },
  "splits": {
    "NuminaMath-CoT": "train",
    "OpenR1-Math-220k": "train",
    "GSM8K": "train",
    "NuminaMath-TIR": "train",
    "OpenThoughts-114k": "train",
    "Bespoke-Stratos-17k": "train",
    "turkish-math-186k": "train",
    "gsm8k_tr": "train",
    "OpenThoughts-TR-18k": "train",
    "turkish-general-reasoning-28k": "train"
  },
  "produced_tokens_by_source_approx": {
    "NuminaMath-CoT": 107190793,
    "OpenR1-Math-220k": 95497705,
    "GSM8K": 17011778,
    "NuminaMath-TIR": 36813446,
    "OpenThoughts-114k": 68179019,
    "Bespoke-Stratos-17k": 20667220,
    "turkish-math-186k": 87527993,
    "gsm8k_tr": 17031848,
    "OpenThoughts-TR-18k": 37179184,
    "turkish-general-reasoning-28k": 12904920
  },
  "examples_by_source_approx": {
    "NuminaMath-CoT": 215040,
    "OpenR1-Math-220k": 16384,
    "turkish-math-186k": 159744,
    "OpenThoughts-114k": 8192,
    "NuminaMath-TIR": 49152,
    "OpenThoughts-TR-18k": 10240,
    "GSM8K": 86016,
    "Bespoke-Stratos-17k": 5411,
    "gsm8k_tr": 83968,
    "turkish-general-reasoning-28k": 6144
  },
  "skipped_too_long_by_source": {},
  "note": "Very long examples are truncated inside tokenizer at max_example_tokens for speed.",
  "disabled_sources_during_this_run": {},
  "active_sources_at_end": [
    "NuminaMath-CoT",
    "OpenR1-Math-220k",
    "GSM8K",
    "NuminaMath-TIR",
    "OpenThoughts-114k",
    "Bespoke-Stratos-17k",
    "turkish-math-186k",
    "gsm8k_tr",
    "OpenThoughts-TR-18k",
    "turkish-general-reasoning-28k"
  ]
}

Source mix

Source	Weight	Repo	Split(s)
NuminaMath-CoT	22	`AI-MO/NuminaMath-CoT`	`train`
OpenR1-Math-220k	18	`open-r1/OpenR1-Math-220k`	`train`
GSM8K	4	`openai/gsm8k`	`train`
NuminaMath-TIR	8	`AI-MO/NuminaMath-TIR`	`train`
OpenThoughts-114k	12	`open-thoughts/OpenThoughts-114k`	`train`
Bespoke-Stratos-17k	4	`bespokelabs/Bespoke-Stratos-17k`	`train`
turkish-math-186k	18	`ituperceptron/turkish-math-186k`	`train`
gsm8k_tr	4	`ytu-ce-cosmos/gsm8k_tr`	`train`
OpenThoughts-TR-18k	7	`selimc/OpenThoughts-TR-18k`	`train`
turkish-general-reasoning-28k	3	`ituperceptron/turkish-general-reasoning-28k`	`train`

License note

This is a mixed-source tokenized derivative dataset. Check each upstream dataset license/provenance before public or commercial use.

Downloads last month: 66