Hugging Face's logo

Datasets:
Khushdholi
/
SDG_Cosmopedia_medium

Modalities:
Text
Formats:
parquet
Size:
< 1K
ArXiv:
Tags:
Synthetic
distilabel
rlaif
Libraries:
Datasets
pandas
Dataset card Viewer Files Community
1
Khushdholi commited on
Commit
1c2e550
•
1 Parent(s): 434ce02

Upload pipeline.yaml with huggingface_hub

Browse files
Files changed (1) hide show
  1. pipeline.yaml +496 -0
pipeline.yaml ADDED
@@ -0,0 +1,496 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ distilabel:
2
+ version: 1.3.2
3
+ pipeline:
4
+ name: synthetic-data-with-llama3.1
5
+ description: null
6
+ steps:
7
+ - step:
8
+ name: load_data_from_hub_0
9
+ resources:
10
+ replicas: 1
11
+ cpus: null
12
+ gpus: null
13
+ memory: null
14
+ resources: null
15
+ input_mappings: {}
16
+ output_mappings: {}
17
+ batch_size: 50
18
+ repo_id: Khushdholi/Cosmopedia_medium
19
+ split: train
20
+ config: null
21
+ streaming: false
22
+ num_examples: null
23
+ storage_options: null
24
+ runtime_parameters_info:
25
+ - name: resources
26
+ runtime_parameters_info:
27
+ - name: replicas
28
+ optional: true
29
+ description: The number of replicas for the step.
30
+ - name: cpus
31
+ optional: true
32
+ description: The number of CPUs assigned to each step replica.
33
+ - name: gpus
34
+ optional: true
35
+ description: The number of GPUs assigned to each step replica.
36
+ - name: memory
37
+ optional: true
38
+ description: The memory in bytes required for each step replica.
39
+ - name: resources
40
+ optional: true
41
+ description: A dictionary containing names of custom resources and the number
42
+ of those resources required for each step replica.
43
+ - name: batch_size
44
+ optional: true
45
+ description: The number of rows that will contain the batches generated by
46
+ the step.
47
+ - name: repo_id
48
+ optional: false
49
+ description: The Hugging Face Hub repository ID of the dataset to load.
50
+ - name: split
51
+ optional: true
52
+ description: The split of the dataset to load. Defaults to 'train'.
53
+ - name: config
54
+ optional: true
55
+ description: The configuration of the dataset to load. This is optional and
56
+ only needed if the dataset has multiple configurations.
57
+ - name: streaming
58
+ optional: true
59
+ description: Whether to load the dataset in streaming mode or not. Defaults
60
+ to False.
61
+ - name: num_examples
62
+ optional: true
63
+ description: The number of examples to load from the dataset. By default will
64
+ load all examples.
65
+ type_info:
66
+ module: distilabel.steps.generators.huggingface
67
+ name: LoadDataFromHub
68
+ name: load_data_from_hub_0
69
+ - step:
70
+ name: text_generation_0
71
+ resources:
72
+ replicas: 1
73
+ cpus: null
74
+ gpus: null
75
+ memory: null
76
+ resources: null
77
+ input_mappings: {}
78
+ output_mappings: {}
79
+ input_batch_size: 50
80
+ llm:
81
+ use_magpie_template: false
82
+ magpie_pre_query_template: null
83
+ generation_kwargs:
84
+ max_new_tokens: 1024
85
+ temperature: 0.7
86
+ model_id: null
87
+ endpoint_name: null
88
+ endpoint_namespace: null
89
+ base_url: http://172.18.30.8:80
90
+ tokenizer_id: null
91
+ model_display_name: llama8B
92
+ structured_output: null
93
+ type_info:
94
+ module: distilabel.llms.huggingface.inference_endpoints
95
+ name: InferenceEndpointsLLM
96
+ group_generations: false
97
+ add_raw_output: true
98
+ num_generations: 1
99
+ use_system_prompt: true
100
+ runtime_parameters_info:
101
+ - name: resources
102
+ runtime_parameters_info:
103
+ - name: replicas
104
+ optional: true
105
+ description: The number of replicas for the step.
106
+ - name: cpus
107
+ optional: true
108
+ description: The number of CPUs assigned to each step replica.
109
+ - name: gpus
110
+ optional: true
111
+ description: The number of GPUs assigned to each step replica.
112
+ - name: memory
113
+ optional: true
114
+ description: The memory in bytes required for each step replica.
115
+ - name: resources
116
+ optional: true
117
+ description: A dictionary containing names of custom resources and the number
118
+ of those resources required for each step replica.
119
+ - name: input_batch_size
120
+ optional: true
121
+ description: The number of rows that will contain the batches processed by
122
+ the step.
123
+ - name: llm
124
+ runtime_parameters_info:
125
+ - name: generation_kwargs
126
+ description: The kwargs to be propagated to either `generate` or `agenerate`
127
+ methods within each `LLM`.
128
+ keys:
129
+ - name: max_new_tokens
130
+ optional: true
131
+ - name: frequency_penalty
132
+ optional: true
133
+ - name: logit_bias
134
+ optional: true
135
+ - name: presence_penalty
136
+ optional: true
137
+ - name: seed
138
+ optional: true
139
+ - name: stop_sequences
140
+ optional: true
141
+ - name: temperature
142
+ optional: true
143
+ - name: tool_choice
144
+ optional: true
145
+ - name: tool_prompt
146
+ optional: true
147
+ - name: tools
148
+ optional: true
149
+ - name: top_p
150
+ optional: true
151
+ - name: do_sample
152
+ optional: true
153
+ - name: repetition_penalty
154
+ optional: true
155
+ - name: return_full_text
156
+ optional: true
157
+ - name: top_k
158
+ optional: true
159
+ - name: typical_p
160
+ optional: true
161
+ - name: watermark
162
+ optional: true
163
+ - name: endpoint_name
164
+ optional: true
165
+ description: The name of the Inference Endpoint to use for the LLM.
166
+ - name: endpoint_namespace
167
+ optional: true
168
+ description: The namespace of the Inference Endpoint to use for the LLM.
169
+ - name: base_url
170
+ optional: true
171
+ description: The base URL to use for the Inference Endpoints API requests.
172
+ - name: api_key
173
+ optional: true
174
+ description: The API key to authenticate the requests to the Inference Endpoints
175
+ API.
176
+ - name: structured_output
177
+ optional: true
178
+ description: The structured output format to use across all the generations.
179
+ - name: add_raw_output
180
+ optional: true
181
+ description: Whether to include the raw output of the LLM in the key `raw_output_<TASK_NAME>`
182
+ of the `distilabel_metadata` dictionary output column
183
+ - name: num_generations
184
+ optional: true
185
+ description: The number of generations to be produced per input.
186
+ type_info:
187
+ module: distilabel.steps.tasks.text_generation
188
+ name: TextGeneration
189
+ name: text_generation_0
190
+ - step:
191
+ name: text_generation_1
192
+ resources:
193
+ replicas: 1
194
+ cpus: null
195
+ gpus: null
196
+ memory: null
197
+ resources: null
198
+ input_mappings: {}
199
+ output_mappings: {}
200
+ input_batch_size: 50
201
+ llm:
202
+ use_magpie_template: false
203
+ magpie_pre_query_template: null
204
+ generation_kwargs:
205
+ max_new_tokens: 1024
206
+ temperature: 0.7
207
+ model_id: null
208
+ endpoint_name: null
209
+ endpoint_namespace: null
210
+ base_url: http://172.18.30.2:80
211
+ tokenizer_id: null
212
+ model_display_name: llama405B
213
+ structured_output: null
214
+ type_info:
215
+ module: distilabel.llms.huggingface.inference_endpoints
216
+ name: InferenceEndpointsLLM
217
+ group_generations: false
218
+ add_raw_output: true
219
+ num_generations: 1
220
+ use_system_prompt: true
221
+ runtime_parameters_info:
222
+ - name: resources
223
+ runtime_parameters_info:
224
+ - name: replicas
225
+ optional: true
226
+ description: The number of replicas for the step.
227
+ - name: cpus
228
+ optional: true
229
+ description: The number of CPUs assigned to each step replica.
230
+ - name: gpus
231
+ optional: true
232
+ description: The number of GPUs assigned to each step replica.
233
+ - name: memory
234
+ optional: true
235
+ description: The memory in bytes required for each step replica.
236
+ - name: resources
237
+ optional: true
238
+ description: A dictionary containing names of custom resources and the number
239
+ of those resources required for each step replica.
240
+ - name: input_batch_size
241
+ optional: true
242
+ description: The number of rows that will contain the batches processed by
243
+ the step.
244
+ - name: llm
245
+ runtime_parameters_info:
246
+ - name: generation_kwargs
247
+ description: The kwargs to be propagated to either `generate` or `agenerate`
248
+ methods within each `LLM`.
249
+ keys:
250
+ - name: max_new_tokens
251
+ optional: true
252
+ - name: frequency_penalty
253
+ optional: true
254
+ - name: logit_bias
255
+ optional: true
256
+ - name: presence_penalty
257
+ optional: true
258
+ - name: seed
259
+ optional: true
260
+ - name: stop_sequences
261
+ optional: true
262
+ - name: temperature
263
+ optional: true
264
+ - name: tool_choice
265
+ optional: true
266
+ - name: tool_prompt
267
+ optional: true
268
+ - name: tools
269
+ optional: true
270
+ - name: top_p
271
+ optional: true
272
+ - name: do_sample
273
+ optional: true
274
+ - name: repetition_penalty
275
+ optional: true
276
+ - name: return_full_text
277
+ optional: true
278
+ - name: top_k
279
+ optional: true
280
+ - name: typical_p
281
+ optional: true
282
+ - name: watermark
283
+ optional: true
284
+ - name: endpoint_name
285
+ optional: true
286
+ description: The name of the Inference Endpoint to use for the LLM.
287
+ - name: endpoint_namespace
288
+ optional: true
289
+ description: The namespace of the Inference Endpoint to use for the LLM.
290
+ - name: base_url
291
+ optional: true
292
+ description: The base URL to use for the Inference Endpoints API requests.
293
+ - name: api_key
294
+ optional: true
295
+ description: The API key to authenticate the requests to the Inference Endpoints
296
+ API.
297
+ - name: structured_output
298
+ optional: true
299
+ description: The structured output format to use across all the generations.
300
+ - name: add_raw_output
301
+ optional: true
302
+ description: Whether to include the raw output of the LLM in the key `raw_output_<TASK_NAME>`
303
+ of the `distilabel_metadata` dictionary output column
304
+ - name: num_generations
305
+ optional: true
306
+ description: The number of generations to be produced per input.
307
+ type_info:
308
+ module: distilabel.steps.tasks.text_generation
309
+ name: TextGeneration
310
+ name: text_generation_1
311
+ - step:
312
+ name: group_columns_0
313
+ resources:
314
+ replicas: 1
315
+ cpus: null
316
+ gpus: null
317
+ memory: null
318
+ resources: null
319
+ input_mappings: {}
320
+ output_mappings: {}
321
+ input_batch_size: 50
322
+ columns:
323
+ '0': generation
324
+ '1': model_name
325
+ output_columns:
326
+ '0': generations
327
+ '1': model_names
328
+ runtime_parameters_info:
329
+ - name: resources
330
+ runtime_parameters_info:
331
+ - name: replicas
332
+ optional: true
333
+ description: The number of replicas for the step.
334
+ - name: cpus
335
+ optional: true
336
+ description: The number of CPUs assigned to each step replica.
337
+ - name: gpus
338
+ optional: true
339
+ description: The number of GPUs assigned to each step replica.
340
+ - name: memory
341
+ optional: true
342
+ description: The memory in bytes required for each step replica.
343
+ - name: resources
344
+ optional: true
345
+ description: A dictionary containing names of custom resources and the number
346
+ of those resources required for each step replica.
347
+ - name: input_batch_size
348
+ optional: true
349
+ description: The number of rows that will contain the batches processed by
350
+ the step.
351
+ type_info:
352
+ module: distilabel.steps.columns.group
353
+ name: GroupColumns
354
+ name: group_columns_0
355
+ - step:
356
+ name: ultra_feedback_0
357
+ resources:
358
+ replicas: 1
359
+ cpus: null
360
+ gpus: null
361
+ memory: null
362
+ resources: null
363
+ input_mappings: {}
364
+ output_mappings: {}
365
+ input_batch_size: 50
366
+ llm:
367
+ use_magpie_template: false
368
+ magpie_pre_query_template: null
369
+ generation_kwargs:
370
+ max_new_tokens: 1024
371
+ temperature: 0.7
372
+ model_id: null
373
+ endpoint_name: null
374
+ endpoint_namespace: null
375
+ base_url: http://172.18.30.2:80
376
+ tokenizer_id: null
377
+ model_display_name: llama405B
378
+ structured_output: null
379
+ type_info:
380
+ module: distilabel.llms.huggingface.inference_endpoints
381
+ name: InferenceEndpointsLLM
382
+ group_generations: false
383
+ add_raw_output: true
384
+ num_generations: 1
385
+ aspect: overall-rating
386
+ runtime_parameters_info:
387
+ - name: resources
388
+ runtime_parameters_info:
389
+ - name: replicas
390
+ optional: true
391
+ description: The number of replicas for the step.
392
+ - name: cpus
393
+ optional: true
394
+ description: The number of CPUs assigned to each step replica.
395
+ - name: gpus
396
+ optional: true
397
+ description: The number of GPUs assigned to each step replica.
398
+ - name: memory
399
+ optional: true
400
+ description: The memory in bytes required for each step replica.
401
+ - name: resources
402
+ optional: true
403
+ description: A dictionary containing names of custom resources and the number
404
+ of those resources required for each step replica.
405
+ - name: input_batch_size
406
+ optional: true
407
+ description: The number of rows that will contain the batches processed by
408
+ the step.
409
+ - name: llm
410
+ runtime_parameters_info:
411
+ - name: generation_kwargs
412
+ description: The kwargs to be propagated to either `generate` or `agenerate`
413
+ methods within each `LLM`.
414
+ keys:
415
+ - name: max_new_tokens
416
+ optional: true
417
+ - name: frequency_penalty
418
+ optional: true
419
+ - name: logit_bias
420
+ optional: true
421
+ - name: presence_penalty
422
+ optional: true
423
+ - name: seed
424
+ optional: true
425
+ - name: stop_sequences
426
+ optional: true
427
+ - name: temperature
428
+ optional: true
429
+ - name: tool_choice
430
+ optional: true
431
+ - name: tool_prompt
432
+ optional: true
433
+ - name: tools
434
+ optional: true
435
+ - name: top_p
436
+ optional: true
437
+ - name: do_sample
438
+ optional: true
439
+ - name: repetition_penalty
440
+ optional: true
441
+ - name: return_full_text
442
+ optional: true
443
+ - name: top_k
444
+ optional: true
445
+ - name: typical_p
446
+ optional: true
447
+ - name: watermark
448
+ optional: true
449
+ - name: endpoint_name
450
+ optional: true
451
+ description: The name of the Inference Endpoint to use for the LLM.
452
+ - name: endpoint_namespace
453
+ optional: true
454
+ description: The namespace of the Inference Endpoint to use for the LLM.
455
+ - name: base_url
456
+ optional: true
457
+ description: The base URL to use for the Inference Endpoints API requests.
458
+ - name: api_key
459
+ optional: true
460
+ description: The API key to authenticate the requests to the Inference Endpoints
461
+ API.
462
+ - name: structured_output
463
+ optional: true
464
+ description: The structured output format to use across all the generations.
465
+ - name: add_raw_output
466
+ optional: true
467
+ description: Whether to include the raw output of the LLM in the key `raw_output_<TASK_NAME>`
468
+ of the `distilabel_metadata` dictionary output column
469
+ - name: num_generations
470
+ optional: true
471
+ description: The number of generations to be produced per input.
472
+ type_info:
473
+ module: distilabel.steps.tasks.ultrafeedback
474
+ name: UltraFeedback
475
+ name: ultra_feedback_0
476
+ connections:
477
+ - from: load_data_from_hub_0
478
+ to:
479
+ - text_generation_0
480
+ - text_generation_1
481
+ - from: text_generation_0
482
+ to:
483
+ - group_columns_0
484
+ - from: text_generation_1
485
+ to:
486
+ - group_columns_0
487
+ - from: group_columns_0
488
+ to:
489
+ - ultra_feedback_0
490
+ - from: ultra_feedback_0
491
+ to: []
492
+ routing_batch_functions: []
493
+ type_info:
494
+ module: distilabel.pipeline.local
495
+ name: Pipeline
496
+ requirements: []