You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: packages/tasks/src/tasks/common-definitions.json
+30-19Lines changed: 30 additions & 19 deletions
Original file line number
Diff line number
Diff line change
@@ -32,7 +32,10 @@
32
32
"description": "The corresponding probability."
33
33
}
34
34
},
35
-
"required": ["label", "score"]
35
+
"required": [
36
+
"label",
37
+
"score"
38
+
]
36
39
},
37
40
"GenerationParameters": {
38
41
"title": "GenerationParameters",
@@ -43,67 +46,75 @@
43
46
"type": "number",
44
47
"description": "The value used to modulate the next token probabilities."
45
48
},
46
-
"topK": {
49
+
"top_k": {
47
50
"type": "integer",
48
51
"description": "The number of highest probability vocabulary tokens to keep for top-k-filtering."
49
52
},
50
-
"topP": {
53
+
"top_p": {
51
54
"type": "number",
52
55
"description": "If set to float < 1, only the smallest set of most probable tokens with probabilities that add up to top_p or higher are kept for generation."
53
56
},
54
-
"typicalP": {
57
+
"typical_p": {
55
58
"type": "number",
56
59
"description": " Local typicality measures how similar the conditional probability of predicting a target token next is to the expected conditional probability of predicting a random token next, given the partial text already generated. If set to float < 1, the smallest set of the most locally typical tokens with probabilities that add up to typical_p or higher are kept for generation. See [this paper](https://hf.co/papers/2202.00666) for more details."
57
60
},
58
-
"epsilonCutoff": {
61
+
"epsilon_cutoff": {
59
62
"type": "number",
60
63
"description": "If set to float strictly between 0 and 1, only tokens with a conditional probability greater than epsilon_cutoff will be sampled. In the paper, suggested values range from 3e-4 to 9e-4, depending on the size of the model. See [Truncation Sampling as Language Model Desmoothing](https://hf.co/papers/2210.15191) for more details."
61
64
},
62
-
"etaCutoff": {
65
+
"eta_cutoff": {
63
66
"type": "number",
64
67
"description": "Eta sampling is a hybrid of locally typical sampling and epsilon sampling. If set to float strictly between 0 and 1, a token is only considered if it is greater than either eta_cutoff or sqrt(eta_cutoff) * exp(-entropy(softmax(next_token_logits))). The latter term is intuitively the expected next token probability, scaled by sqrt(eta_cutoff). In the paper, suggested values range from 3e-4 to 2e-3, depending on the size of the model. See [Truncation Sampling as Language Model Desmoothing](https://hf.co/papers/2210.15191) for more details."
65
68
},
66
-
"maxLength": {
69
+
"max_length": {
67
70
"type": "integer",
68
71
"description": "The maximum length (in tokens) of the generated text, including the input."
69
72
},
70
-
"maxNewTokens": {
73
+
"max_new_tokens": {
71
74
"type": "integer",
72
75
"description": "The maximum number of tokens to generate. Takes precedence over maxLength."
73
76
},
74
-
"minLength": {
77
+
"min_length": {
75
78
"type": "integer",
76
79
"description": "The minimum length (in tokens) of the generated text, including the input."
77
80
},
78
-
"minNewTokens": {
81
+
"min_new_tokens": {
79
82
"type": "integer",
80
83
"description": "The minimum number of tokens to generate. Takes precedence over maxLength."
81
84
},
82
-
"doSample": {
85
+
"do_sample": {
83
86
"type": "boolean",
84
87
"description": "Whether to use sampling instead of greedy decoding when generating new tokens."
85
88
},
86
-
"earlyStopping": {
89
+
"early_stopping": {
87
90
"description": "Controls the stopping condition for beam-based methods.",
"description": "Number of beams to use for beam search."
93
104
},
94
-
"numBeamGroups": {
105
+
"num_beam_groups": {
95
106
"type": "integer",
96
107
"description": "Number of groups to divide num_beams into in order to ensure diversity among different groups of beams. See [this paper](https://hf.co/papers/1610.02424) for more details."
97
108
},
98
-
"penaltyAlpha": {
109
+
"penalty_alpha": {
99
110
"type": "number",
100
111
"description": "The value balances the model confidence and the degeneration penalty in contrastive search decoding."
101
112
},
102
-
"useCache": {
113
+
"use_cache": {
103
114
"type": "boolean",
104
115
"description": "Whether the model should use the past last key/values attentions to speed up decoding"
0 commit comments