Datasets:

shrenikb
/

specrouter-hidden-states

Tasks:

Text Generation

Tags:

speculative-decoding

routing

llm-inference

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed

Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
contract: struct<concatenated_width: int64, contract_id: string, hidden_width: int64, layer_indices: list<item (... 50 chars omitted)
  child 0, concatenated_width: int64
  child 1, contract_id: string
  child 2, hidden_width: int64
  child 3, layer_indices: list<item: int64>
      child 0, item: int64
  child 4, layer_semantics: string
  child 5, version: int64
dtype: string
files: struct<hidden.npy: struct<sha256: string, size_bytes: int64>, index.jsonl: struct<sha256: string, si (... 17 chars omitted)
  child 0, hidden.npy: struct<sha256: string, size_bytes: int64>
      child 0, sha256: string
      child 1, size_bytes: int64
  child 1, index.jsonl: struct<sha256: string, size_bytes: int64>
      child 0, sha256: string
      child 1, size_bytes: int64
format_version: int64
index_file: string
metadata: struct<base_model: string, counterfactual: string, cycles: string, cycles_sha256: string, eagle_mode (... 49 chars omitted)
  child 0, base_model: string
  child 1, counterfactual: string
  child 2, cycles: string
  child 3, cycles_sha256: string
  child 4, eagle_model: string
  child 5, traces: string
  child 6, traces_sha256: string
positions: int64
requests: int64
shape: list<item: int64>
  child 0, item: int64
source: struct<path: string, sha256: string, size_bytes: int64>
  child 0, path: string
  child 1, sha256: string
  child 2, size_bytes: int64
trajectory_set_sha256: string
dataset: string
request_id: string
start: int64
split: string
count: int64
trajectory_sha256: string
to
{'count': Value('int64'), 'dataset': Value('string'), 'request_id': Value('string'), 'split': Value('string'), 'start': Value('int64'), 'trajectory_sha256': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              contract: struct<concatenated_width: int64, contract_id: string, hidden_width: int64, layer_indices: list<item (... 50 chars omitted)
                child 0, concatenated_width: int64
                child 1, contract_id: string
                child 2, hidden_width: int64
                child 3, layer_indices: list<item: int64>
                    child 0, item: int64
                child 4, layer_semantics: string
                child 5, version: int64
              dtype: string
              files: struct<hidden.npy: struct<sha256: string, size_bytes: int64>, index.jsonl: struct<sha256: string, si (... 17 chars omitted)
                child 0, hidden.npy: struct<sha256: string, size_bytes: int64>
                    child 0, sha256: string
                    child 1, size_bytes: int64
                child 1, index.jsonl: struct<sha256: string, size_bytes: int64>
                    child 0, sha256: string
                    child 1, size_bytes: int64
              format_version: int64
              index_file: string
              metadata: struct<base_model: string, counterfactual: string, cycles: string, cycles_sha256: string, eagle_mode (... 49 chars omitted)
                child 0, base_model: string
                child 1, counterfactual: string
                child 2, cycles: string
                child 3, cycles_sha256: string
                child 4, eagle_model: string
                child 5, traces: string
                child 6, traces_sha256: string
              positions: int64
              requests: int64
              shape: list<item: int64>
                child 0, item: int64
              source: struct<path: string, sha256: string, size_bytes: int64>
                child 0, path: string
                child 1, sha256: string
                child 2, size_bytes: int64
              trajectory_set_sha256: string
              dataset: string
              request_id: string
              start: int64
              split: string
              count: int64
              trajectory_sha256: string
              to
              {'count': Value('int64'), 'dataset': Value('string'), 'request_id': Value('string'), 'split': Value('string'), 'start': Value('int64'), 'trajectory_sha256': Value('string')}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1880, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

count int64	dataset string	request_id string	split string	start int64	trajectory_sha256 string
96	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0000	train	0	4169e1b6b6ff0b5a0650e81525b3f3a12299a06ea422e7c310f12f0843b0e116
95	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0001	train	96	87b100dcbb82338699593f93107451f9a48d4461bb1acecbd85b9861293e4db8
103	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0002	train	191	0d6d3619122c5f972ff3983a56dd212c2999cd544ba4ea6fc411cd0847de5666
86	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0003	train	294	5ada9025d7671c5802c6727e48c96013d3ac34101b33dd2f6256eca7a62a0463
50	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0004	train	380	fbcb7ebdff27cf212d372323308d1ed848b30bbad9d038521ed4ecc263247300
102	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0005	train	430	050c1ded63610276262f1f87cdc5408ad9b23f9a304cb0b62d0e67d0ced005ab
48	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0006	train	532	6ce998c40107a4b32ff00d0138baf80e01602e0f2aad0175a8f43562fa6ebe03
82	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0007	train	580	2c59ac5ac6ddad0644786cb4c33089bf7bb884a0d766627fe64c399ad2f1f3ce
160	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0008	train	662	1cfa8cd9af3c4edfc43fc0f07a94056fad4d67d2eeb63bebbcc836af5785079f
136	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0009	train	822	fe42f4a1d523143d0f54275f9c992e57304ed3e15e4d22a2d37b79341b26354a
78	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0010	train	958	8b07e4787f33efd5cf740a2f82469c9e57a1eb730b1a08772c66b4f8969a6489
74	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0011	train	1,036	76f53674191006ecb97ff38cf52827b5066d837c7f751f8d5b2587633575f455
49	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0012	train	1,110	a2e973f3bff46bc0bc585af94ba745e2b91d18142da54e024d2d979698a9545b
128	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0013	train	1,159	eb83c575a5b0d6ad8034544ba9d45e18fb7a3e0d724139f033ce8de833e7a72b
97	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0014	train	1,287	365c47f7e0e24fa356d3a11bd7c6c507bc7f9e2a58c37dbc9f5922ddf84bb028
32	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0015	train	1,384	17a1125fee28e8e5a4789f6d14d7e708104a7fa83e49d634265a4de32882d1c3
60	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0016	train	1,416	4fe6a1ba4d85f5a25b84782e342d033fc495daebe0c928a69c859973e298d479
70	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0017	train	1,476	1b91680fb5825a726909c8149930bc6fefaf11c9c7c401e4b012e9c87004a4e4
84	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0018	train	1,546	f4dbade3472ec53673d5a2fa2e45d65f908bfdcfad1e643084f41599ebc26b6b
81	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0019	train	1,630	bd3a5e541c7a24a866dac2f66d6b5b2194cd88bb12b304d84ddab0afd1f8848d
24	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0020	train	1,711	c9e702477af6ca6ca6007cb6d23cd5cb6dba94a927434e47791b264034405407
77	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0021	train	1,735	c2e7be9cc7386da0637b1cffee72eb1edce39a949afa26d3c8f29f8912c3cb43
85	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0022	train	1,812	b2e6f979c6eb89514adfc98a673c9722bcc8c997a521e9bbdb42f758643b9180
87	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0023	train	1,897	06b5ab01214a5ed74dd99341bd36a57424688089408b4970ea1ed12ff13fc951
71	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0024	train	1,984	2b2f85f9fe7807fa6c39d7e5a8edaddfae7259369926a0f4ace4aef739252602
109	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0025	train	2,055	4693097f87ee4c45822012fd660c496b410497cdb6a8e12d57877f08d45a49c4
71	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0026	train	2,164	da5feaf1c0caaebfc1ff0559558db6559358138fbbe1a2cd5e72308a9fb8be67
63	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0027	train	2,235	a817d3e82147e3c3fcf7937301b7a4a17cc90b6641fa88e225c8a9600ff9726a
44	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0028	train	2,298	4ab6d250c165a2e907f02134275fbe6207effc94f1b70b181984b4424b2b0598
46	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0029	train	2,342	58cee6cd26b706a975e420f001764891e48ddb0f60fbf14d20de408c31d6cdd6
256	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0030	train	2,388	64bdf4daf3d06b479dc59eaa322e9d569deace6486e392b7fb281a84b1d8b2a8
51	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0031	train	2,644	5e6eb8c1daccc423f4c699335b8525dcc2fa8de2987b2f0014ca8175c684018e
62	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0032	train	2,695	b1a71edca829cb54af74455494d9c1c0f711e5d95413544542db901f47de4d87
114	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0033	train	2,757	0bd941240425195bfb76080ccd2074593b8d833796886b5a6ea5c18f51f50a46
256	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0034	train	2,871	5b8bab6c4359a2f9601181556e1e661d4dfbed28de3f85e761caa69e3a99c6da
76	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0035	train	3,127	7bf33703346d209e71612cfa3023be123c69072c33f7e94679ad372b693fcae0
72	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0036	train	3,203	811407e28d7fdd3f6d08c6245b102f2260e88f7e235e56222b636ff09e3d5cb7
42	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0037	train	3,275	934aef89b8e6a0c3f867cc269cd56b9f8889152d86804cf69e16a79342685e7f
71	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0038	train	3,317	1b72fde5b468e581b2da8b2453c31169060a0b147ef96f08ef7cfb75ddf99586
59	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0039	train	3,388	1e4456ac61f699a613f2bc50780af4cdd94a7694ec53b243bdf7e319c204cb00
56	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0040	train	3,447	7e2d7f956dec029283909de8f41552dad63735a49e69e37ff52b54ac09e96ca9
29	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0041	train	3,503	0a2bab250343697c15aea1f944391170f0ef2b9a378c83b4270ea718c8f1f385
74	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0042	train	3,532	be681cd08956aa4cc152262f3f1a4453dce4fe8fbd373084742094c16d1ddea2
60	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0043	train	3,606	6477cc97ea7c79fd1f707b9fb163a37b7cb6788e2b8f0abeabd1b734939b05cb
76	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0044	train	3,666	101ee963e65864a87af42fedbfeed0ff86de4ae3f3bd7cdddfcbdaf775b25e52
25	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0045	train	3,742	9b5cc34d21a7425bbcc8e3b3abb4d495f5490d00013782317d4576c0bdc52e4d
147	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0046	train	3,767	1cad929441e6ede2c3574f744ae15ee04c8bcf2feb322ab1bc2da28cc28b2e79
64	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0047	train	3,914	add85f941acc46a62bcff634b4be82e93a00c98e106a26eb998278befc21648d
60	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0048	train	3,978	fb494b2a82b6da072d191a200b097d743e9add09a4800398fe0d002236c34ac7
98	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0049	train	4,038	86133ab5aa75eb89887db7dd05f56d2a587c775d7bc5161aac61b361f1291aeb
71	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0050	train	4,136	83d93283f118f212f7e40cf0a2754f379006ef38e7b17ea6c996d2d9d6c352c9
27	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0051	train	4,207	8714296e69b81df97c60a4bd4e63195a7135a3ac394786827c34f533184f3748
39	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0052	train	4,234	b12c57c3d5ae0c6a7b6411032da52fdaabf40c811219d4d13339e18774408e7b
256	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0053	train	4,273	c7606adda6b79d1115f78ef6071bcc0d71a78d7255c50ff62d6afb2daabbba01
98	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0054	train	4,529	a7ed64d48872ec250acf7b9941193333a5f79a37b15b878a10200b5f323273e4
103	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0055	train	4,627	fe6493d7c2a3d5d5ba6e5da593ade1396316962e13b2a743b0deb94c24c7bd82
64	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0056	train	4,730	a36050978b01bd5ec39dd6a1350a593e46926cf4be8c0b5f35bc04bc2d5ea956
69	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0057	train	4,794	d71100c9577101b418190ea49e06ae65bc2c7aeec00925d2c881f4adb823b1e8
92	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0058	train	4,863	a2a064169d93cd5eece59280a0a5bedade605bc42b0dc42313ef157c927f0513
94	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0059	train	4,955	32961bba9f39b5dae437657b57d6978bd1f019798bb1ce20464b5e1dc8990c11
60	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0060	train	5,049	ba217c03c16425e693cf109fc61347a7f885a05aaeef63edd68bc0943eeff68d
41	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0061	train	5,109	115e5a9bfabf7aa8fa367118787ad9a320313d6735aed0dea657456301b7fb9b
256	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0062	train	5,150	c468df0aea54f186af7b00f278a6c8e7c91e3c706754e4e795bf2160f2896ed8
62	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0063	train	5,406	6a762e6ea8a42f71f788d15633ee4507b05db172298059b249f2576bfa309b5e
125	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0064	train	5,468	6a4372053b96fab6cfc49554057f985c3fd211523a9daf2a54d89592cf786d1d
52	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0065	train	5,593	4f5b3f697de000d8cf9c271fc2d09fd8eb8dcfa0017abf043cc40e99d77045f4
64	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0066	train	5,645	3b7454f4e9aac78c683295e9d876e78f9e4178ad6d61d6f7c3616cfc0abac5f9
100	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0067	train	5,709	68ad9fd2f782509b5b34944b8afcd9ac883b6374f6e940091e7476a740c2ea2b
102	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0068	train	5,809	1f336711decb4b3451bb72b07138d7fc0371a895c996786049d019fb7635148f
122	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0069	train	5,911	28c5ffd500315575288b4ffef16d336f74ac84aeb0d2a1aa7421fb6285073603
131	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0070	train	6,033	fb58bde657b054414002ae8a1f5a3492151f8b983f92cb9b3bcd73e43e7981e6
49	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0071	train	6,164	8406cd5f03988e2629916dc74db8f8298818ee6a57eab19f7148708acf4c47cb
208	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0072	train	6,213	51cffbf69aa84f768d4b8f0fd55cc601aa06f4b6ea7cf02e65dc5167c0588c77
35	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0073	train	6,421	a463ffc9c1f8e8986c2df7a58269595293aab9d55566e2f4138d60cc0998092c
182	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0074	train	6,456	06672d93dd8555f446935a47200027353f17d9313f82e026d9fe017d0e4d0f5b
81	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0075	train	6,638	3018ddbfc0bb90e5666d7b6150c09da83dbf5589b1df9a6b15a9f2319b041aab
70	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0076	train	6,719	07deb76f8e14e2a60edff4a348dff9f256e6d284bb5cfff2fe3543f11b89004a
20	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0077	train	6,789	a2ee21afa42f53e94370546654b4ba4441f3ba279944b5c4c9020cce087c1469
30	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0078	train	6,809	f67625a58a81420c741c019dbade14fff8241432769b700575adf070f48a1397
153	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0079	train	6,839	56d6dc870bd76865dd5bb97c94c0fbad76603803f806cb74ad3d171ca38c6d6e
83	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0080	train	6,992	76133faecf5f9eab851f3ff5e6520793de61fb5f3e0b88aab1ce10cd1112a893
63	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0081	train	7,075	d4a09a1b2392a93f82dcbc1820173b0e9c803a34ac1afc95d88db7ef71eaba29
90	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0082	train	7,138	c8933a8d93e874e0b4d64d66981cc86bde7664e1546b8675ef11c759725ad3b7
91	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0083	train	7,228	f04ac005628747fed32975766f157c13f6c1cf71d7b601ff14ae441536ccd152
38	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0084	train	7,319	7dd61af73946215af1441dccb4bf4990534f487b9c8573e8787fe8bbca460a51
48	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0085	train	7,357	b3fc606392b79dba8a40a549d09dae109b877919de1b01ad3fb157b2df2749c7
154	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0086	train	7,405	8fb5deaeee65e8324b531b0e6467568092433b18a333f9ad0404ddf22707e979
21	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0087	train	7,559	33a033fe2d8e9021b394466606054a1fbe4aafc8e6731f887744425fab15ccfc
51	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0088	train	7,580	a8a2ad1b1fa83385c86dcdfbcfa42316e1cb84c7596f5aa55b2624746a0b1093
84	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0089	train	7,631	6c9378632a0773183ea70c1502d0daa17b4c10d398822cbae20341e3b2f7238c
65	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0090	train	7,715	08647f25a4ebdd8de822e70506d4bf7017ff614d1cc81dcda181ba4ee91cb0b0
70	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0091	train	7,780	8c32be8c396cbed708fb73f18332e46ad6fac535be88390c9833951fc34e856a
82	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0092	train	7,850	2728051d94fa47bfd94aa8ff482dc53f06adfad1b270f528610262dc9eb8f08b
25	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0093	train	7,932	b2ecaf94758a78aba7958a0773548d27edb7fdb29765fba168d76a0fb2f0b0a9
44	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0094	train	7,957	f13de44d5e0660e9ceb68ef2ff86798b405329e7a9dd197bebdf81b3a3598109
56	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0095	train	8,001	d510a9cd93b313c92705397ae38cc4951f01b26f37903fe36307deeb2afea6be
59	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0096	train	8,057	82baacfbff074c3de19746dfef0259dd770fa123aa76a2fda5c7472dc2ed71d5
154	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0097	train	8,116	18deff4256dee00c72a68b57e746b9378de2a67d19b17278e5d11e1866aac951
30	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0098	train	8,270	41e4012a23eae4293f5b90e21188a8bda6813c4ed35828f37788251a022f1c86
72	hvd-2wikimultihopqa	hvd-2wikimultihopqa:hvd-2wikimultihopqa-train-0099	train	8,300	1c7bc7554e5f5b62c5e984c0974cc4341d90114d580c3122f13273bc25d1ad6e

End of preview.

SpecRouter target hidden states

Router input features for SpecRouter, which learns a cost-aware per-block router over heterogeneous speculative decoding mechanisms (no-spec, ngram, suffix, EAGLE-3, DFlash).

These are only needed to train routers. Every other experiment -- oracle ladders, exact-DP economics, cost tables, action-subset grids, DFlash and EAGLE comparisons -- runs entirely from the data bundle committed in the GitHub repo.

Each cell is one target model on one workload:

<campaign>/<target>/<dataset>/
    hidden.npy      # [positions, layers, hidden_width] float16
    index.jsonl     # per-request row offsets
    manifest.json   # contract, shape, source hashes

The tensors follow the eagle3-target-layer-inputs contract: decoder-layer inputs at three sampled depths, concatenated per position.

Campaigns

Key	Meaning
`v2`	Complete 32-cell campaign backing the four-action published result
`v3-bf16`	Corrected shared-BF16 rebuild; partial, see the repo ledger

Do not mix campaigns. They use different target execution and different EAGLE tree configurations. MASTER_RESULTS.md in the GitHub repo is the authoritative ledger for which artifacts are valid.

Usage

git clone https://github.com/shrenikbhansali/SpecRouter.git && cd SpecRouter
python implementations/specrouter/scripts/setup_data.py

# list available cells
python implementations/specrouter/scripts/fetch_hidden_states.py

# fetch selectively
python implementations/specrouter/scripts/fetch_hidden_states.py \
  --campaign v3-bf16 --target qwen3-8b --download

Files land under their normal campaign paths, so training scripts work unchanged. Fetch verifies every file against the SHA-256 recorded in the cell's own manifest.

Downloads last month: 124

Total file size:

94.6 GB

SpecRouter target hidden states

Contents

Campaigns

Usage