nics-efc/TaH2-1.7B-max2
Text Generation • 2B • Updated • 190
real_text large_stringlengths 1.2k 85.6k | real_token large listlengths 263 16.4k | mask large listlengths 263 16.4k | data_id int64 0 123k |
|---|---|---|---|
<|im_start|>user
If $ω > 0$, the graph of the function $y=\cos (ωx+ \frac {π}{3})$ is shifted to the right by $\frac {π}{3}$ units and coincides with the graph of the function $y=\sin ωx$, then the minimum value of $ω$ is ( ).
A: $ \frac {11}{2}$
B: $ \frac {5}{2}$
C: $ \frac {1}{2}$
D: $ \frac {3}{2}$
Please reason ... | [
151644,
872,
198,
2679,
400,
56871,
861,
220,
15,
54876,
279,
4771,
315,
279,
729,
400,
88,
34433,
9407,
320,
56871,
87,
10,
1124,
37018,
314,
48245,
15170,
18,
5410,
3,
374,
28973,
311,
279,
1290,
553,
57960,
37018,
314,
48245,
15170... | [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0... | 0 |
<|im_start|>user
Please solve the programming task below in Python. Code should be wrapped in a markdown code block.
def logistic_robot(n: int, c: int, commands: List[str]) -> List[int]:
"""A logistics company uses an automated delivery robot to distribute packages along a linear series of warehouses.
Args:
... | [
151644,
872,
198,
5501,
11625,
279,
15473,
3383,
3685,
304,
13027,
13,
6119,
1265,
387,
19472,
304,
264,
50494,
2038,
2504,
382,
750,
71710,
59116,
1445,
25,
526,
11,
272,
25,
526,
11,
11293,
25,
1759,
17303,
2467,
1464,
1759,
18640,
... | [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0... | 1 |
"<|im_start|>user\nLet \\( a, b, c, d, e, f \\) be integers selected from the set \\( \\{1,2, \\ldot(...TRUNCATED) | [151644,872,198,10061,17767,264,11,293,11,272,11,294,11,384,11,282,1124,8,387,25780,4091,504,279,738(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 2 |
"<|im_start|>user\nPlease solve the programming task below in Python. Code should be wrapped in a ma(...TRUNCATED) | [151644,872,198,5501,11625,279,15473,3383,3685,304,13027,13,6119,1265,387,19472,304,264,50494,2038,2(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 3 |
"<|im_start|>user\nPlease solve the programming task below in Python. Code should be wrapped in a ma(...TRUNCATED) | [151644,872,198,5501,11625,279,15473,3383,3685,304,13027,13,6119,1265,387,19472,304,264,50494,2038,2(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 4 |
"<|im_start|>user\nPlease solve the programming task below in Python. Code should be wrapped in a ma(...TRUNCATED) | [151644,872,198,5501,11625,279,15473,3383,3685,304,13027,13,6119,1265,387,19472,304,264,50494,2038,2(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 5 |
"<|im_start|>user\nthe average temperature for tuesday , wednesday and thursday was 52 ° c . the av(...TRUNCATED) | [151644,872,198,1782,5461,9315,369,259,7047,1154,10840,7316,323,270,7181,572,220,20,17,36286,272,659(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 6 |
"<|im_start|>user\nA book store had 4 books in the bargain bin. They sold 3 books, but then added mo(...TRUNCATED) | [151644,872,198,32,2311,3553,1030,220,19,6467,304,279,44563,9544,13,2379,6088,220,18,6467,11,714,122(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 7 |
"<|im_start|>user\nWhat phenomenon causes the backward curling of fluid as it exits a nozzle?A: Bern(...TRUNCATED) | [151644,872,198,3838,24844,11137,279,27555,13957,287,315,15590,438,432,42086,264,81881,30,32,25,1416(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 8 |
"<|im_start|>user\nA school is planning to upgrade its heating system from an old oil-based system t(...TRUNCATED) | [151644,872,198,32,2906,374,9115,311,13910,1181,23579,1849,504,458,2310,5590,5980,1849,311,264,803,1(...TRUNCATED) | [0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0(...TRUNCATED) | 9 |
End of preview. Expand in Data Studio
TaH2 AMTeam Tool
Code: thu-nics/TaH · Paper: arXiv:2609.35748.
Final training data for Qwen3-1.7B, 4B, and 8B: math, code, science, and tool calling.
| Student | Directory | Train rows | Train tokens | Eval rows |
|---|---|---|---|---|
| 1.7B | 1.7b/ |
273,195 | 1,099,413,406 | 955 |
| 4B | 4b/ |
638,817 | 2,586,857,725 | 1,000 |
| 8B | 8b/ |
1,282,124 | 5,192,783,145 | 1,000 |
Each directory contains train/, eval/, and mix_report.json.
Splits are pretokenized Hugging Face save_to_disk datasets with fields
real_text, real_token, mask, and data_id. Only assistant tokens have mask=1.
The 1.7B responses were regenerated by Qwen3-8B, with Qwen3-32B simulating tool responses. The 4B/8B data keeps original AM teacher responses and Nemotron tool traces.
The 1.7B generation inputs are in 1.7b-prompts/{am,tool_calling}/, with
train.jsonl and eval.jsonl in each: AM has 220,523 / 811 prompts; tool calling
has 51,073 / 189 full reference traces. These precede episode expansion and subsampling.
from huggingface_hub import snapshot_download
from datasets import load_from_disk
snapshot_download("nics-efc/TaH2-amteam-tool", repo_type="dataset",
local_dir="data", allow_patterns=["1.7b/*"])
train = load_from_disk("data/1.7b/train")
Replace 1.7b/* with 4b/*, 8b/*, or 1.7b-prompts/* to download another subset.
- Downloads last month
- 74
Models trained or fine-tuned on nics-efc/TaH2-amteam-tool
