Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
13
16
domain
stringclasses
5 values
src
stringclasses
150 values
source_dataset
stringclasses
11 values
source_config
stringclasses
3 values
source_split
stringclasses
2 values
source_revision
stringclasses
11 values
source_row
int64
0
859k
⌀
source_ref
stringlengths
121
233
thinking
bool
2 classes
sampling
stringclasses
2 values
seed
int64
2.41k
1.07B
max_tokens
int32
2.05k
4.1k
finish_reason
stringclasses
2 values
n_prompt_tokens
int32
12
3.07k
n_output_tokens
int32
2
4.1k
prompt_ids
listlengths
12
3.07k
output_ids
listlengths
2
4.1k
prompt_text
stringlengths
74
15.9k
output_text
stringlengths
11
31.1k
train-math-4741
math
metamath
meta-math/MetaMathQA
null
train
aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18
208,857
{"dataset": "meta-math/MetaMathQA", "revision": "aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18", "split": "train", "row": 208857}
false
liquid
883,487,038
2,048
stop
51
149
[ 124894, 124899, 5922, 207, 2882, 2021, 1997, 3487, 310, 6025, 2480, 64005, 229, 1845, 9055, 302, 51431, 467, 48341, 309, 1188, 10852, 835, 462, 229, 32, 10074, 20, 1119, 355, 278, 1579, 302, 9055, 302, 51431, 395, 1188, 3487, 310, 4657,...
[ 9196, 1997, 677, 2531, 48341, 342, 873, 34, 555, 10099, 17, 310, 13596, 33753, 229, 1845, 9055, 302, 51431, 22, 2298, 10852, 3321, 462, 229, 32, 34, 555, 10074, 63908, 9077, 20, 1500, 278, 2925, 893, 2660, 487, 229, 32, 34, 555, 100...
<|startoftext|><|im_start|>user If Carla needs to dry-clean 80 pieces of laundry by noon and she starts work at 8 AM, what is the number of pieces of laundry that she needs to clean per hour?<|im_end|> <|im_start|>assistant <think></think>
Carla has until noon (12:00 PM) to finish drying 80 pieces of laundry. She starts working at 8:00 AM. First, find the total time available from 8:00 AM to 12:00 PM: - From 8:00 AM to 12:00 PM is 4 hours. Next, divide the total number of pieces of laundry by the total time in hours to find out how many pieces she need...
train-multi-722
multi
wildchat-Portuguese
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "d51f86ec270067924761041b28db3066", "turn_identifier": 1234438}
false
greedy
672,087,481
2,048
stop
42
349
[ 124894, 124899, 5922, 207, 369, 3214, 10607, 2077, 17044, 2188, 1576, 1257, 7726, 301, 662, 17044, 112886, 10251, 1761, 10619, 3880, 320, 277, 425, 1176, 20, 29197, 7726, 301, 662, 17044, 89271, 405, 41267, 39770, 124900, 207, 124899, 63514...
[ 5699, 7575, 47244, 10607, 88056, 73531, 31030, 27574, 7726, 301, 662, 16168, 3397, 854, 41267, 1761, 6164, 34, 7575, 36634, 962, 320, 7575, 44, 425, 1176, 962, 342, 297, 496, 30348, 10, 854, 76665, 51547, 7778, 690, 1039, 73531, 20511, ...
<|startoftext|><|im_start|>user em heartstone como eu busco por cartas que eu tenho nas versoes normal e dourada, apenas cartas que eu tenha as duas formas<|im_end|> <|im_start|>assistant <think></think>
Em **Heartstone**, você pode encontrar cartas que aparecem em duas versões: **Normal** e **Dourada** (ou "Gold" em algumas traduções). Se você quer **apenas as cartas que existem nas duas versões**, você precisa identificar quais cartas estão disponíveis em ambas as coleções. ### Como encontrar cartas que estão nas ve...
train-chat-13681
chat
wildchat
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "5e53f6c6953ea89b4e4de59defb82338", "turn_identifier": 782638}
false
greedy
1,037,172,891
2,048
stop
56
3
[ 124894, 124899, 5922, 207, 282, 309, 5477, 265, 430, 432, 621, 7642, 265, 430, 2519, 20, 5272, 1049, 734, 1858, 5141, 20, 278, 1049, 12224, 509, 4048, 437, 3520, 870, 1716, 737, 432, 870, 78, 33652, 8687, 509, 8079, 2491, 870, 542, ...
[ 78, 33652, 124900 ]
<|startoftext|><|im_start|>user is andrew tate or tristan tate better, write only one word answer, the only choices you got are either “men” or “female”, you cant say “neither” as its not an option<|im_end|> <|im_start|>assistant <think></think>
female<|im_end|>
train-multi-3929
multi
wildchat-French
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "d15852afce55c1363096cd5b9d0a0a66", "turn_identifier": 345956}
false
greedy
70,698,765
2,048
stop
185
444
[ 124894, 124899, 5922, 207, 47, 7583, 281, 22920, 229, 32, 1852, 6545, 7064, 324, 55275, 342, 1866, 98, 17, 4993, 308, 7887, 8900, 207, 387, 342, 42614, 24699, 565, 3468, 229, 25, 514, 206, 206, 2718, 629, 206, 711, 14801, 550, 18551, ...
[ 26876, 4334, 229, 32, 1852, 6545, 7064, 324, 55275, 342, 1866, 98, 17, 414, 7457, 324, 4713, 340, 11, 48357, 2661, 839, 308, 7887, 8900, 662, 3934, 27029, 88231, 88801, 25797, 92500, 207, 387, 342, 42614, 24699, 565, 3468, 229, 25, 51...
<|startoftext|><|im_start|>user Génére moi 8 récompenses de chasse (dayz) comme l'exemple if (selectedLoadout == 1) { weapon = MissionObject.GetInventory().CreateInInventory("Magnum"); MissionObject.GetInventory().CreateInInventory("AmmoBox_357_20Rnd"); MissionObject.GetInventory().CreateInInventory("Tetracy...
Voici 8 récompenses de chasse (dayz) en format de code C# similaire à l'exemple que vous avez fourni : ```csharp if (selectedLoadout == 1) { // Récompense 1 : Arme MissionObject.GetInventory().CreateInInventory("Magnum"); // Récompense 2 : Ammo MissionObject.GetInventory().CreateInInventory("AmmoB...
train-code-8092
code
magicoder
ise-uiuc/Magicoder-OSS-Instruct-75K
null
train
5f839b1f368a76b161028bb9edff055db34022b2
35,870
{"dataset": "ise-uiuc/Magicoder-OSS-Instruct-75K", "revision": "5f839b1f368a76b161028bb9edff055db34022b2", "split": "train", "row": 35870, "index": 5723}
false
liquid
138,565,829
2,048
stop
316
515
[ 124894, 124899, 5922, 207, 3582, 437, 60889, 415, 22938, 267, 1446, 395, 116100, 278, 4739, 302, 278, 3570, 96278, 72, 1446, 296, 52016, 22, 440, 3570, 96278, 72, 1446, 355, 1092, 310, 23737, 267, 10972, 1446, 1219, 267, 2496, 10904, 29...
[ 2596, 4528, 278, 3570, 95972, 5977, 29324, 72, 1446, 405, 4524, 20, 522, 1094, 310, 1360, 1017, 6940, 63990, 25, 22, 92143, 278, 10904, 60509, 1409, 64380, 487, 278, 3570, 15410, 72, 23357, 318, 26, 22, 10704, 787, 278, 20505, 10904, ...
<|startoftext|><|im_start|>user You are tasked with implementing a function that simulates the behavior of the `setTimeout` function in JavaScript. The `setTimeout` function is used to execute a specified function after a given delay in milliseconds. Your task is to create a Python function that takes in a session obje...
To implement the `ExecuteSetTimeout` function as described, we need to follow these steps: 1. Extract the delay (`ms`) from the `params` dictionary. 2. Check if the extracted delay is a float. If not, raise a `ValueError`. 3. Use the `time.sleep()` function to wait for the specified number of milliseconds (delay). 4. ...
train-code-3995
code
magicoder
ise-uiuc/Magicoder-OSS-Instruct-75K
null
train
5f839b1f368a76b161028bb9edff055db34022b2
43,315
{"dataset": "ise-uiuc/Magicoder-OSS-Instruct-75K", "revision": "5f839b1f368a76b161028bb9edff055db34022b2", "split": "train", "row": 43315, "index": 608}
false
greedy
128,987,504
2,048
stop
152
614
[ 124894, 124899, 5922, 207, 3582, 437, 60889, 415, 22938, 267, 4388, 49381, 1177, 1492, 340, 7239, 395, 510, 9435, 33446, 17655, 302, 9502, 22, 440, 1177, 1252, 373, 2721, 310, 23737, 267, 10972, 1579, 302, 9502, 487, 4820, 33942, 309, 5...
[ 2596, 4528, 278, 3570, 14235, 59, 50324, 72, 1014, 296, 340, 7239, 395, 510, 9435, 33446, 17655, 302, 9502, 20, 522, 510, 1016, 278, 3570, 14952, 1639, 28172, 72, 309, 3570, 14952, 1639, 97842, 72, 310, 9686, 278, 17655, 302, 9502, 22...
<|startoftext|><|im_start|>user You are tasked with implementing a task scheduling system using C++ that can handle concurrent execution of tasks. The system should be able to execute a specified number of tasks from multiple threads and ensure that the tasks are executed correctly. Your task is to create a class `Tas...
To implement the `TaskScheduler` class in C++ that can handle concurrent execution of tasks, we can use the `std::thread` and `std::mutex` to manage the execution of tasks. The class will have two methods: `schedule` to add a task and `execute` to run all scheduled tasks concurrently. Here's the implementation: ```cpp...
train-math-553
math
metamath
meta-math/MetaMathQA
null
train
aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18
358,711
{"dataset": "meta-math/MetaMathQA", "revision": "aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18", "split": "train", "row": 358711}
true
liquid
179,868,570
4,096
stop
70
827
[ 124894, 124899, 5922, 207, 15386, 355, 4686, 5435, 20, 374, 2552, 4433, 1394, 1784, 22, 904, 7590, 310, 3260, 709, 893, 8036, 383, 4686, 5435, 19270, 20, 467, 229, 28, 4433, 22, 4409, 10966, 310, 9809, 229, 632, 893, 1512, 734, 2524, ...
[ 2083, 1094, 310, 22415, 278, 2719, 34, 496, 15386, 355, 4686, 5435, 20, 374, 2552, 4433, 1394, 1784, 22, 904, 7590, 310, 3260, 709, 893, 8036, 383, 4686, 5435, 19270, 20, 467, 229, 28, 4433, 22, 4409, 10966, 310, 9809, 229, 632, 893...
<|startoftext|><|im_start|>user Mark is reading books, for x hours each day. He decided to increase his time spent on reading books weekly, by 4 hours. Mark wants to spend 18 time during one week on reading books. What is the value of unknown variable x? What is the value of unknown variable x?<|im_end|> <|im_start|>as...
We need to parse the problem: "Mark is reading books, for x hours each day. He decided to increase his time spent on reading books weekly, by 4 hours. Mark wants to spend 18 time during one week on reading books. What is the value of unknown variable x?" Interpretation: Mark reads for x hours per day. Over a week (7 d...
train-tool-2353
tool
hermes-toolresult
NousResearch/hermes-function-calling-v1
func_calling
train
dae3e1d28cfbcf4b915c04ea1e072030529b4bda
453
{"dataset": "NousResearch/hermes-function-calling-v1", "revision": "dae3e1d28cfbcf4b915c04ea1e072030529b4bda", "split": "train", "row": 453, "config": "func_calling", "id": "4b944828-6815-4a98-8689-b4c417220807", "cut_messages": 5}
true
greedy
696,002,077
4,096
stop
1,264
510
[ 124894, 124899, 23630, 207, 125017, 207, 38184, 2554, 6380, 496, 18631, 71, 53003, 2289, 4781, 42917, 97704, 1377, 496, 19021, 6380, 496, 18522, 91, 278, 11430, 2628, 302, 267, 9551, 374, 267, 3179, 13107, 296, 278, 11310, 53, 1177, 50007...
[ 2083, 1094, 310, 6119, 310, 278, 4695, 54447, 278, 1703, 1446, 9795, 551, 880, 19770, 11058, 22, 440, 6189, 2485, 1413, 612, 1703, 5913, 6943, 2556, 12144, 22, 1275, 1252, 47275, 278, 9079, 8369, 63908, 7731, 589, 8385, 2868, 14330, 487...
<|startoftext|><|im_start|>system <tools> {"name": "track_vehicle_purchase_history", "description": "Tracks the purchase history of a vehicle for a specific customer in the CRM system.", "parameters": {"type": "object", "properties": {"customer_id": {"type": "string", "description": "The unique identifier of the custom...
We need to respond to the user confirming the three function calls have been executed successfully. The tool results show all three functions returned success messages. We should summarize the outcomes clearly. Let's extract key info from each result: 1. track_vehicle_purchase_history: success, purchase_date 2023-03-...
train-math-2934
math
metamath
meta-math/MetaMathQA
null
train
aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18
72,197
{"dataset": "meta-math/MetaMathQA", "revision": "aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18", "split": "train", "row": 72197}
false
liquid
211,065,186
2,048
stop
50
541
[ 124894, 124899, 5922, 207, 26671, 267, 13643, 302, 278, 11142, 395, 5173, 36632, 27, 1510, 229, 26, 81, 12, 310, 36632, 31, 640, 723, 3370, 1119, 355, 278, 3620, 1579, 395, 532, 13643, 5173, 36632, 28, 1510, 229, 29, 81, 12, 310, 39...
[ 2083, 437, 2496, 267, 13643, 302, 278, 11142, 395, 11606, 278, 1628, 9062, 310, 278, 3620, 1579, 91622, 27, 19, 26, 81, 47935, 310, 278, 1628, 9062, 310, 91622, 31, 22306, 89782, 334, 13643, 296, 278, 3620, 11142, 510, 373, 9744, 467,...
<|startoftext|><|im_start|>user Given a translation of the plane that takes $-3 + 2i$ to $-7 - i$, what is the complex number that this translation takes $-4 + 5i$ to?<|im_end|> <|im_start|>assistant <think></think>
We are given a translation of the plane that maps the point corresponding to the complex number \(-3+2i\) to the point corresponding to \(-7-i\). A translation in the complex plane can be represented by adding a fixed complex number \(z_0\) to any complex number. That is, if we denote the translation as \(T(z) = z + w\...
train-math-5442
math
gsm8k-train
openai/gsm8k
main
train
740312add88f781978c0658806c59bc2815b9866
1,119
{"dataset": "openai/gsm8k", "revision": "740312add88f781978c0658806c59bc2815b9866", "split": "train", "row": 1119, "config": "main"}
true
liquid
657,959,649
4,096
stop
101
1,111
[ 124894, 124899, 5922, 207, 50, 436, 47381, 893, 1472, 1784, 10288, 310, 949, 63494, 383, 278, 8660, 22, 229, 440, 1008, 1784, 302, 278, 2524, 1188, 47381, 229, 29, 4611, 383, 278, 8660, 415, 949, 63494, 22, 229, 7634, 302, 278, 2272, ...
[ 2083, 1094, 310, 11994, 34, 496, 50, 436, 47381, 893, 1472, 1784, 10288, 310, 949, 63494, 383, 278, 8660, 22, 440, 1008, 1784, 302, 278, 2524, 1188, 47381, 229, 29, 4611, 383, 278, 8660, 415, 949, 63494, 22, 7634, 302, 278, 2272, 28...
<|startoftext|><|im_start|>user Jill spends time every day talking to her boyfriend on the phone. The first day of the week she spends 5 minutes on the phone with her boyfriend. Each of the following days she spends twice as much time talking on the phone as the day before. After the 5th day her parents take the phon...
We need to solve: "Jill spends time every day talking to her boyfriend on the phone. The first day of the week she spends 5 minutes on the phone with her boyfriend. Each of the following days she spends twice as much time talking on the phone as the day before. After the 5th day her parents take the phone away from her...
train-chat-2036
chat
wildchat
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "74f6fcdfeabb9516a63723c14deaf9ed", "turn_identifier": 345893}
true
liquid
996,265,764
4,096
stop
26
1,265
[ 124894, 124899, 5922, 207, 22497, 267, 13483, 1194, 267, 1994, 20589, 267, 2600, 661, 58036, 10981, 927, 21072, 3451, 38, 124900, 207, 124899, 63514, 207, 124901 ]
[ 2083, 1094, 310, 5272, 267, 13483, 1184, 267, 1994, 309, 267, 2600, 661, 58036, 13508, 22, 440, 4695, 27472, 1400, 10972, 278, 5175, 3451, 20, 600, 522, 510, 8830, 734, 432, 2951, 374, 124394, 39, 1938, 278, 10996, 2997, 34, 496, 2249...
<|startoftext|><|im_start|>user Write a conversation where a human asks a superintelligent ai <some question><|im_end|> <|im_start|>assistant <think>
We need to write a conversation between a human and a superintelligent AI. The user hasn't specified the exact question, so we can invent one or ask for clarification? But the instruction says: "Write a conversation where a human asks a superintelligent ai <some question>". So I should pick a plausible question that a ...
train-multi-4955
multi
wildchat-Chinese
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "75fb83aefd5eeabbb1dafe1a6c75b194", "turn_identifier": 1682747}
true
greedy
772,528,884
4,096
stop
23
1,339
[ 124894, 124899, 5922, 207, 31025, 4466, 13812, 1608, 20740, 37218, 41137, 49843, 42762, 55533, 4637, 5132, 70913, 124900, 207, 124899, 63514, 207, 124901 ]
[ 597, 4695, 20589, 34, 496, 31025, 4466, 13812, 1608, 20740, 37218, 41137, 49843, 42762, 55533, 4637, 5132, 70913, 10, 643, 41842, 310, 496, 32721, 622, 5272, 267, 20219, 383, 1226, 11464, 1219, 7226, 41388, 4110, 2426, 1672, 685, 1976, 26...
<|startoftext|><|im_start|>user 帮我写一份关于教师资格培训后的心得感想<|im_end|> <|im_start|>assistant <think>
The user asks: "帮我写一份关于教师资格培训后的心得感想" which translates to "Help me write a reflection on my feelings after teacher qualification training." So they want a piece of writing, presumably a personal reflection on the experience of teacher qualification training. We need to produce a well-structured essay or paragraph(s) abo...
train-multi-1407
multi
aya-arb
CohereForAI/aya_dataset
null
train
f9ea04583f02a8f86404ff6c58bf75fe637df8a2
100,183
{"dataset": "CohereForAI/aya_dataset", "revision": "f9ea04583f02a8f86404ff6c58bf75fe637df8a2", "split": "train", "row": 100183}
false
liquid
458,732,081
2,048
stop
28
478
[ 124894, 124899, 5922, 207, 30142, 67999, 85472, 76909, 100358, 5079, 104720, 101698, 6418, 35699, 70571, 3439, 70376, 3220, 1539, 5079, 104720, 124900, 207, 124899, 63514, 207, 124901, 124902 ]
[ 102969, 3279, 55063, 101519, 24050, 68384, 59278, 3623, 70981, 100358, 5079, 104720, 342, 69053, 2436, 21803, 792, 3279, 8815, 35857, 9415, 3220, 3279, 5039, 67905, 17, 21691, 14575, 70155, 38580, 5878, 69643, 2774, 21691, 1570, 6075, 43292, ...
<|startoftext|><|im_start|>user تحاول دوائر صناعة التبغ القضاء على عمل الأطفال في مزارع التبغ<|im_end|> <|im_start|>assistant <think></think>
نعم، هناك جهود متزايدة من شركات صناعة التبغ (مثل باكو، روكستار، إلخ) ومؤسسات غير حكومية ومبادرات دولية للحد من عمالة الأطفال في مزارع التبغ. إليك بعض النقاط الرئيسية: ### 1. **التزامات قانونية** - **قوانين العمل:** تلتزم معظم شركات التبغ بقوانين العمل الدولية، بما في ذلك حظر عمالة الأطفال في المزارع. في العديد من ا...
train-tool-1917
tool
hermes-toolresult
NousResearch/hermes-function-calling-v1
func_calling
train
dae3e1d28cfbcf4b915c04ea1e072030529b4bda
127
{"dataset": "NousResearch/hermes-function-calling-v1", "revision": "dae3e1d28cfbcf4b915c04ea1e072030529b4bda", "split": "train", "row": 127, "config": "func_calling", "id": "33e8c808-952e-4e3c-ad17-5e492313e0ed", "cut_messages": 3}
false
greedy
88,358,102
2,048
stop
426
169
[ 124894, 124899, 23630, 207, 125017, 207, 38184, 2554, 6380, 496, 92882, 13118, 3328, 71, 39655, 1377, 496, 19021, 6380, 496, 87969, 27930, 487, 267, 4572, 3848, 20, 1951, 11866, 5061, 309, 893, 20, 3848, 2931, 20, 309, 7457, 50007, 496, ...
[ 597, 20505, 27930, 374, 278, 4572, 3848, 496, 32030, 71, 1118, 530, 94651, 3663, 498, 10, 355, 405, 6852, 63990, 21, 7575, 99455, 13386, 309, 8260, 65505, 229, 1217, 27, 21, 5634, 21, 885, 60, 1035, 34, 1163, 34, 555, 66, 207, 21, ...
<|startoftext|><|im_start|>system <tools> {"name": "extract_video_metadata", "description": "Extract metadata from a video file, including capture date and time, file size, and format.", "parameters": {"type": "object", "properties": {"file_path": {"type": "string", "description": "The file path to the video file from ...
The extracted metadata for the video file "family_beach_day.mov" is as follows: - **Capture Date and Time**: 2023-03-15T14:30:00Z - **File Size**: 110,100,482 bytes (approximately 105MB) - **File Format**: MOV - **Additional Details**: - Duration: 00:01:42 - Frame Rate: 29.97 fps - Resolution: 1920x1080 - Audi...
train-code-3365
code
magicoder
ise-uiuc/Magicoder-OSS-Instruct-75K
null
train
5f839b1f368a76b161028bb9edff055db34022b2
36,543
{"dataset": "ise-uiuc/Magicoder-OSS-Instruct-75K", "revision": "5f839b1f368a76b161028bb9edff055db34022b2", "split": "train", "row": 36543, "index": 4704}
false
greedy
752,808,601
2,048
stop
241
551
[ 124894, 124899, 5922, 207, 3582, 437, 60889, 415, 22938, 267, 4648, 3294, 1317, 3918, 296, 47300, 22, 440, 3294, 1252, 373, 27948, 20, 5063, 435, 4759, 1558, 278, 1654, 309, 1263, 3895, 22, 5565, 4388, 355, 310, 2996, 267, 2339, 3570, ...
[ 2596, 11994, 532, 2719, 20, 522, 1094, 310, 4528, 267, 4648, 3294, 1317, 3918, 296, 47300, 2087, 3570, 19461, 785, 29944, 49060, 969, 2339, 1252, 6833, 278, 1654, 309, 1263, 3895, 405, 92155, 309, 2675, 267, 2010, 310, 3498, 787, 267, ...
<|startoftext|><|im_start|>user You are tasked with implementing a custom range data structure in Rust. The range should be inclusive, meaning it includes both the start and end values. Your task is to create a struct `RangeInclusive` that stores the start and end values as pointers and provides a method to check if a ...
To solve this problem, we need to implement a custom range data structure in Rust called `RangeInclusive`. This struct should store the start and end values as pointers and provide a method to check if a given value falls within the range (inclusive of both start and end values). ### Approach 1. **Struct Definition**:...
train-chat-1066
chat
ultrachat
HuggingFaceH4/ultrachat_200k
null
train_sft
8049631c405ae6576f93f445c6b8166f76f5505a
28,703
{"dataset": "HuggingFaceH4/ultrachat_200k", "revision": "8049631c405ae6576f93f445c6b8166f76f5505a", "split": "train_sft", "row": 28703, "prompt_id": "48a0b84a5df423ece0fa7d3eefa7274fd9ab732e92499cadfdb91b96bcb131a3"}
false
liquid
101,713,882
2,048
stop
116
1,032
[ 124894, 124899, 5922, 207, 15743, 1022, 21516, 9798, 302, 679, 6661, 20, 5272, 267, 10432, 395, 510, 14253, 8385, 26337, 1317, 487, 267, 5146, 309, 8184, 278, 1702, 18296, 310, 267, 17722, 5450, 627, 9705, 22, 440, 1317, 509, 5339, 1252...
[ 2596, 20277, 532, 4388, 20, 522, 687, 1016, 278, 2272, 21417, 63990, 21, 3570, 8525, 2778, 87984, 1389, 3024, 29199, 19352, 318, 21, 3570, 4184, 1527, 6655, 97561, 72, 342, 5890, 3570, 1805, 28, 72, 2770, 2488, 22415, 18672, 3639, 318, ...
<|startoftext|><|im_start|>user Using any Python library of your choice, write a script that can automatically extract structured data from a website and send the information gathered to a designated Slack channel. The data you collect should be of interest to your team, and the Slack channel should be configured in a ...
To accomplish this task, we will use the following libraries: - `requests`: For making HTTP requests. - `BeautifulSoup` (from `bs4`): To parse HTML content. - `slack_sdk`: To interact with Slack's API. We'll also create a configuration file named `.config` to store sensitive information like API keys and login creden...
train-code-5358
code
evolcode
theblackcat102/evol-codealpaca-v1
null
train
c75242318519b5470635e84064baed7c78594020
3,658
{"dataset": "theblackcat102/evol-codealpaca-v1", "revision": "c75242318519b5470635e84064baed7c78594020", "split": "train", "row": 3658}
false
greedy
897,936,520
2,048
length
74
2,048
[ 124894, 124899, 5922, 207, 63204, 267, 23948, 8599, 432, 10829, 395, 513, 1049, 11663, 738, 1689, 787, 401, 5635, 390, 22795, 46450, 16304, 20, 296, 278, 1033, 302, 267, 4160, 20, 36654, 401, 401, 9594, 302, 267, 91360, 20, 702, 735, ...
[ 597, 4695, 20589, 34, 496, 63204, 267, 23948, 8599, 432, 10829, 395, 513, 1049, 11663, 738, 1689, 787, 401, 5635, 390, 22795, 46450, 16304, 20, 296, 278, 1033, 302, 267, 4160, 20, 36654, 401, 401, 9594, 302, 267, 91360, 20, 702, 735, ...
<|startoftext|><|im_start|>user Construct a computational procedure or algorithm that not only ascertains if an input alphanumeric entity, in the form of a string, constitutes an anagram of a palindrome, but also determines the possible palindromes that can be made with the given string. The resulted palindromes should...
The user asks: "Construct a computational procedure or algorithm that not only ascertains if an input alphanumeric entity, in the form of a string, constitutes an anagram of a palindrome, but also determines the possible palindromes that can be made with the given string. The resulted palindromes should be unique and a...
train-code-7470
code
magicoder
ise-uiuc/Magicoder-OSS-Instruct-75K
null
train
5f839b1f368a76b161028bb9edff055db34022b2
69,362
{"dataset": "ise-uiuc/Magicoder-OSS-Instruct-75K", "revision": "5f839b1f368a76b161028bb9edff055db34022b2", "split": "train", "row": 69362, "index": 22969}
false
greedy
925,356,766
2,048
stop
249
1,661
[ 124894, 124899, 5922, 207, 3582, 437, 2496, 267, 21516, 4713, 88534, 395, 12403, 19549, 1446, 12569, 309, 9795, 267, 1446, 2447, 383, 2796, 3151, 22, 5565, 4388, 355, 310, 2345, 278, 4713, 309, 5191, 278, 9643, 5405, 37762, 467, 91319, ...
[ 2596, 11994, 278, 2719, 20, 2443, 589, 16391, 278, 2496, 4713, 88534, 3300, 467, 3300, 63990, 14066, 13162, 325, 44909, 312, 833, 25797, 63513, 207, 38807, 550, 15872, 44426, 514, 387, 26581, 13291, 4673, 360, 2215, 550, 15872, 13291, 17,...
<|startoftext|><|im_start|>user You are given a Python code snippet that manipulates function arguments and calls a function based on certain conditions. Your task is to understand the code and identify the missing elements denoted by "__" in the code snippet. Additionally, you need to determine the function's behavior...
To solve the problem, let's analyze the given code snippet step by step: ### Code Snippet: ```python giv = len(args) if packed(a): exp = len(a) - 1 args += [__] * (exp - giv) args = args[:exp] + [args[exp:]] else: exp = len(a) args += [__] * (exp - giv) if exp < giv: error("up to %d arg...
train-chat-5996
chat
wildchat
allenai/WildChat-1M
null
train
7d6490e462285cf85d91eabea0f9a954fbddcd1f
null
{"dataset": "allenai/WildChat-1M", "revision": "7d6490e462285cf85d91eabea0f9a954fbddcd1f", "split": "train", "row": null, "conversation_hash": "80cb52be8873c38a10814f4f1d78c390", "turn_identifier": 113502}
false
greedy
766,571,640
2,048
stop
143
1,262
[ 124894, 124899, 5922, 207, 266, 529, 20817, 6669, 38454, 2340, 27, 8193, 4528, 267, 5718, 395, 3926, 19270, 20, 20304, 20, 309, 40200, 24208, 1579, 395, 437, 92395, 487, 267, 12742, 368, 482, 5610, 22, 7645, 1202, 657, 83, 404, 5536, ...
[ 2596, 4528, 267, 5718, 296, 41081, 6669, 4429, 704, 13022, 229, 27, 395, 20471, 19270, 20, 20304, 20, 309, 40200, 24208, 92395, 487, 18731, 368, 482, 1492, 278, 7575, 5536, 657, 83, 404, 5536, 5256, 16275, 1674, 962, 5718, 9798, 20, 1...
<|startoftext|><|im_start|>user In jetpack compose material3 please implement a graph that shows weekly, monthly, and yearly expenses number that are fetched from a firestore collection. Use patrykandpatrick/vico graph library to achieve this implementation "com.patrykandpatrick.vico:core:1.6.5" implementatio...
To implement a graph in Jetpack Compose Material 3 that displays weekly, monthly, and yearly expenses fetched from Firestore using the **patrykandpatrick/vico** graph library, follow these steps: --- ### 1. **Add Dependencies** Ensure the following dependencies are included in your `build.gradle` (Module: app): ```g...
End of preview. Expand in Data Studio

LFM3-1.7B self-distillation set (63M output tokens)

LiquidAI/LFM3-1.7B's own responses to 42,915 prompts from eleven public instruction datasets: chat, code, math, tool calling and multilingual, thinking on and off, greedy and sampled. It is the training and held-out data of the block drafter Lottolabs/LFM3-1.7B-DFlash-Drafter (speculative decoding for the Tenstorrent P150 build Lottolabs/LFM3-1.7B-TT-Mixed-BFP4-BFP8-P150). Every row carries the exact prompt and output token ids the served model saw and produced, the decoded text, the sampling settings and seed, and a pointer to the source row (dataset, revision, split, row / native id).

Not an official Liquid AI release. The responses were generated by Lottolabs with a quantized build of LFM3-1.7B; they are machine-generated and unfiltered (see Limitations).

split rows prompt tokens output tokens finish stop / length
train 42,591 8,989,869 63,346,722 34,909 / 7,682
eval 324 132,879 451,751 263 / 61

Train by domain (rows / output tokens): chat 15,223 / 23.77M, code 8,921 / 17.21M, math 8,656 / 9.96M, multilingual 6,682 / 11.06M, tool 3,109 / 1.35M. Eval: 60 prompts per domain + 24 benchmark rows (domain = "bench", below). Thinking on / off and greedy / liquid sampling are each ~50/50 in every domain.

from datasets import load_dataset
ds = load_dataset("Lottolabs/LFM3-1.7B-SelfDistill-63M")
row = ds["train"][0]
row["prompt_ids"], row["output_ids"]      # token ids (LFM3 tokenizer); prompt_ids end with the generation prompt
row["prompt_text"] + row["output_text"]   # the same, decoded with special tokens

Columns

column type meaning
id string {train,eval}-{domain}-{n} (eval-bench-{workload}-{greedy0,liquid1..3} for bench rows)
domain string chat, code, math, tool, multi (non-English), bench (eval only)
src string source tag: ultrachat, wildchat, wildchat-<Language>, oasst2, oasst2-<lang>, aya-<lang>, magicoder, evolcode, gsm8k-train, metamath, numina, hermes-func_calling_singleturn, hermes-func_calling, hermes-toolresult, glaive; bench: the workload name
source_dataset, source_config, source_split, source_revision string Hugging Face dataset id, config, split and the exact revision the prompt was read from (null for bench rows)
source_row int64 row index in that split at that revision (null for WildChat, which was streamed: use source_ref)
source_ref string (JSON) the above plus the source's native id: UltraChat prompt_id; WildChat conversation_hash + turn_identifier of the first turn; OASST2 message_id; Magicoder index; Hermes id (+ cut_messages for after-tool-result cuts)
thinking bool chat template rendered with enable_thinking (on: the prompt ends with <think>; off: <think></think>)
sampling string greedy (temperature 0, repetition penalty 1.0) or liquid (temperature 0.1, top-k 50, repetition penalty 1.1 = the model's generation_config.json), seeded
seed int64 request seed
max_tokens int32 4,096 with thinking, 2,048 without (bench rows: the workload's limit)
finish_reason string stop (ended with `<
n_prompt_tokens, n_output_tokens int32 lengths of the two id lists
prompt_ids, output_ids list exact token ids sent to and returned by the server
prompt_text, output_text string the ids decoded with the LFM3 tokenizer, special tokens kept (thinking outputs contain the reasoning, then </think> and the answer; tool calls are LFM3's <tool_call> format)

How it was made

Scripts (in scripts/): prep_prompts.py (prompt selection and rendering), gen.py (generation client), make_data.py + data.py (token-id packing used by the drafter training), prep_rowids.py (= prep_prompts.py with dataset revisions pinned and each prompt's source row recorded; its re-run reproduced all 42,591 train prompts byte-identically, then supplied the source_* columns), build_dataset.py (this parquet export; it checks every row's prompt ids against the generation records).

Prompts (prep_prompts.py, random.Random(1234) and the per-dataset shuffle seeds in the script): rendered with LFM3's unmodified chat_template.jinja (generation prompt on; enable_thinking drawn 50/50 per prompt; tool definitions passed as tools), prompts over 3,072 tokens dropped.

  • chat: UltraChat-200k train_sft (25% with up to 3 user turns of context, the conversation cut after a user turn), WildChat-1M English (conversations with any OpenAI-moderation-flagged turn skipped; 20% with 2 user turns), OASST2 English root prompts.
  • multi: WildChat-1M non-English, OASST2 non-English root prompts, Aya (non-English, inputs ≥ 20 characters).
  • code: Magicoder-OSS-Instruct-75K problem, evol-codealpaca-v1 instruction.
  • math: GSM8K main train, MetaMathQA query (de-duplicated by original question), NuminaMath-CoT problem.
  • tool: Hermes function-calling v1 (func_calling_singleturn, func_calling: cut before the first assistant turn, and for multi-turn conversations also after the first tool results, with the dataset's tool calls / results as context), Glaive function-calling v2 (first user turn + the system prompt's function definitions).
  • bench (eval only): the six workloads of the serving repo's speed benchmark (lfm3/tests/http_spec_bench.py: chat_2k = an excerpt of Newton's Opticks (Project Gutenberg, public domain) + a poem request, code_edit / code_think / tool_write = edits of two Python files of the Lottolabs LFM3 runtime source, gsm8k = the first GSM8K test question, chat = a one-line question), each greedy (seed 0) and liquid (seeds 1-3).

Decontamination and splits. The train split uses no UltraChat test_sft, no GSM8K test, and no MetaMathQA row whose original question is a GSM8K test question. A prompt is dropped when its last user turn (whitespace-collapsed, lower-cased, first 300 characters) equals any user turn of the serving repo's quality evaluation texts (UltraChat test, code, multilingual and book prompts) or any earlier-selected prompt — so train and eval share no prompt and no prompt repeats. The eval split is the first 60 kept prompts per domain (+ bench). Not checked against other public benchmarks (e.g. MATH / AIME problems inside NuminaMath-CoT and MetaMathQA, HumanEval-like tasks in the code sets).

Generation (2026-11-13): one Tenstorrent Blackhole P150 serving the GPTQ mixed BFP4/BFP8 checkpoint of Lottolabs/LFM3-1.7B-TT-Mixed-BFP4-BFP8-P150 (release s13, checkpoint/manifest.json SHA256 cf3ba5e4daac03ecf2003ff65a57528bd3819c718eb999d82b428340325e1b2d; base LiquidAI/LFM3-1.7B revision 40ebe0aa6662c9d40389c2f5d565b6a2c6aebacd) with the runtime image dev-s13 (= release s13 lottolabs/lfm3-1.7b-tt-p150:s13-20261112), vLLM OpenAI server, 16 slots, max-model-len 32,768, prompt-lookup speculation on (exact verification: speculation does not change the output). Requests: /v1/completions with the prompt as token ids, return_token_ids, skip_special_tokens: false, per-row seed, max_tokens and sampling as in the columns; 20 concurrent clients (gen.py). Train: 63,346,722 output tokens in 20,308 s (3,119 tok/s aggregate). The outputs are the quantized P150 model's, not BF16 LFM3's (they differ where the two models' top tokens differ); for drafter training that is the point — the drafter learns the served model.

Intended use

Training and evaluating draft models / speculative decoders and other distillation students for LFM3-1.7B (the DFlash drafter trained on train and measured on eval), and studying LFM3-1.7B's output distribution. It is not a curated instruction-tuning set: responses were not checked for correctness or safety.

Limitations

  • Responses are raw LFM3-1.7B outputs (thinking and non-thinking): they can be wrong, repetitive or unsafe; 18% of train rows stopped at max_tokens (finish_reason = "length", mostly long reasoning).
  • Prompts are user-written (WildChat, OASST2, Aya) or synthetic (UltraChat, Magicoder, evol-codealpaca, MetaMathQA rephrasings, Glaive, Hermes — several of them generated with OpenAI models, e.g. Magicoder's card asks users to follow OpenAI's usage policies). WildChat was de-identified by its authors (Presidio + rules) and moderation-flagged conversations were skipped here, but real user text can still contain personal or offensive content.
  • WildChat rows have no row index (the dataset was streamed with a shuffle buffer); use conversation_hash + turn_identifier.

License and attribution

This dataset combines three kinds of material; use it under all of the terms that apply:

  1. Model responses (output_ids, output_text) are outputs of LFM3-1.7B by Liquid AI (quantized by Lottolabs). They are distributed with, and should be treated as subject to, the LFM Open License v1.0 (full text: LICENSE) — notably Section 5: commercial use is licensed only for entities with annual revenue below US $10,000,000. (The license does not address model outputs explicitly; we apply it conservatively.) "LFM" and "Liquid" are Liquid AI's names, used here only to describe the origin of the work.
  2. Prompts are taken (unchanged text, re-rendered with the LFM3 chat template; multi-turn context cut as described above) from the datasets below, under their licenses. Copies: LICENSES/ (Apache-2.0.txt, MIT.txt with the copyright notices, ODC-BY-1.0.md).
  3. The compilation: because it contains a substantial extract of WildChat-1M, this database is a derivative database of WildChat-1M and is made available under the ODC Attribution License (ODC-By) v1.0 (LICENSES/ODC-BY-1.0.md, https://opendatacommons.org/licenses/by/1-0/) as to the database; its individual contents keep the licenses in points 1-2. Contains information from WildChat-1M, which is made available under the ODC Attribution License.
source Hugging Face id revision read license rows train / eval
UltraChat 200k HuggingFaceH4/ultrachat_200k 8049631c405ae6576f93f445c6b8166f76f5505a MIT (© 2023 THUNLP) 8,968 / 28
WildChat-1M allenai/WildChat-1M 7d6490e462285cf85d91eabea0f9a954fbddcd1f ODC-BY 1.0 8,005 / 56
OpenAssistant OASST2 OpenAssistant/oasst2 179dd21fc55192153d94adb0e0ce8f69e222bf75 Apache-2.0 2,468 / 18
Aya Dataset CohereForAI/aya_dataset (now CohereLabs) f9ea04583f02a8f86404ff6c58bf75fe637df8a2 Apache-2.0 2,464 / 18
Magicoder OSS-Instruct 75K ise-uiuc/Magicoder-OSS-Instruct-75K 5f839b1f368a76b161028bb9edff055db34022b2 MIT (© 2023 iSE-UIUC) 4,958 / 37
evol-codealpaca v1 theblackcat102/evol-codealpaca-v1 c75242318519b5470635e84064baed7c78594020 Apache-2.0 3,963 / 23
GSM8K (main, train) openai/gsm8k 740312add88f781978c0658806c59bc2815b9866 MIT (© 2021 OpenAI) 2,848 / 21
MetaMathQA meta-math/MetaMathQA aa4f34d3d2d3231299b5b03d9b3e5a20da45aa18 MIT 2,841 / 21
NuminaMath-CoT AI-MO/NuminaMath-CoT (now project-numina) 9d8d210c9f6a36c8f3cd84045668c9b7800ef517 Apache-2.0 2,967 / 18
Hermes function-calling v1 NousResearch/hermes-function-calling-v1 dae3e1d28cfbcf4b915c04ea1e072030529b4bda Apache-2.0 2,023 / 26
Glaive function-calling v2 glaiveai/glaive-function-calling-v2 e7f4b6456019f5d8bcb991ef0dd67d8ff23221ac Apache-2.0 1,086 / 34

Bench rows (eval, 24): Newton's Opticks (public domain), GSM8K test question 1 (MIT, © 2021 OpenAI), Lottolabs runtime source files (published in the serving repo's runtime/lfm3-runtime-src.tar.gz) and prompts written by Lottolabs. None of the source datasets ships a NOTICE file. Please cite the source datasets' papers (see their cards) when you use this data.

Changes made (Lottolabs): prompts selected, cut to a user turn and rendered with the LFM3 chat template; responses generated with a GPTQ BFP4/BFP8-quantized LFM3-1.7B; packed as token ids with metadata.

Downloads last month
136