lib.rs 34.1 KB
Newer Older
1
pub mod config;
2
mod health;
3
/// Text Generation Inference Webserver
4
mod infer;
5
mod queue;
Olivier Dehaene's avatar
Olivier Dehaene committed
6
pub mod server;
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
7
mod validation;
Olivier Dehaene's avatar
Olivier Dehaene committed
8

9
use infer::{Infer, InferError, InferStreamResponse};
10
use queue::{Entry, Queue};
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
11
use serde::{Deserialize, Serialize};
12
13
use tokio::sync::OwnedSemaphorePermit;
use tokio_stream::wrappers::UnboundedReceiverStream;
14
use utoipa::ToSchema;
Olivier Dehaene's avatar
Olivier Dehaene committed
15
use validation::Validation;
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
16

17
18
19
20
21
22
23
/// Type alias for generation responses
pub(crate) type GenerateStreamResponse = (
    OwnedSemaphorePermit,
    u32, // input_length
    UnboundedReceiverStream<Result<InferStreamResponse, InferError>>,
);

drbh's avatar
drbh committed
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
#[derive(Clone, Deserialize, ToSchema)]
pub(crate) struct VertexInstance {
    #[schema(example = "What is Deep Learning?")]
    pub inputs: String,
    #[schema(nullable = true, default = "null", example = "null")]
    pub parameters: Option<GenerateParameters>,
}

#[derive(Deserialize, ToSchema)]
pub(crate) struct VertexRequest {
    #[serde(rename = "instances")]
    pub instances: Vec<VertexInstance>,
}

#[derive(Clone, Deserialize, ToSchema, Serialize)]
pub(crate) struct VertexResponse {
    pub predictions: Vec<String>,
}

43
44
/// Hub type
#[derive(Clone, Debug, Deserialize)]
45
pub struct HubModelInfo {
46
47
48
49
50
51
    #[serde(rename(deserialize = "id"))]
    pub model_id: String,
    pub sha: Option<String>,
    pub pipeline_tag: Option<String>,
}

52
53
54
55
56
57
58
59
60
61
62
63
64
65
#[derive(Debug, Clone, Deserialize, PartialEq)]
pub struct ChatTemplate {
    name: String,
    template: String,
}

#[derive(Debug, Clone, Deserialize, PartialEq)]
#[serde(untagged)]
pub enum ChatTemplateVersions {
    Single(String),
    Multiple(Vec<ChatTemplate>),
}

#[derive(Debug, Clone, Deserialize, Default)]
66
pub struct HubTokenizerConfig {
67
    pub chat_template: Option<ChatTemplateVersions>,
68
    pub completion_template: Option<String>,
69
    #[serde(deserialize_with = "token_serde::deserialize")]
70
    pub bos_token: Option<String>,
71
    #[serde(deserialize_with = "token_serde::deserialize")]
72
    pub eos_token: Option<String>,
73
74
75
}

impl HubTokenizerConfig {
76
    pub fn from_file(filename: &std::path::Path) -> Self {
77
78
79
80
81
        let content = std::fs::read_to_string(filename).unwrap();
        serde_json::from_str(&content).unwrap_or_default()
    }
}

82
#[derive(Clone, Debug, Deserialize, ToSchema, Serialize)]
drbh's avatar
drbh committed
83
84
#[serde(tag = "type", content = "value")]
pub(crate) enum GrammarType {
85
86
87
88
89
90
91
    /// A string that represents a [JSON Schema](https://json-schema.org/).
    ///
    /// JSON Schema is a declarative language that allows to annotate JSON documents
    /// with types and descriptions.
    #[serde(rename = "json")]
    #[schema(example = json ! ({"properties": {"location":{"type": "string"}}}))]
    Json(serde_json::Value),
drbh's avatar
drbh committed
92
93
94
95
    #[serde(rename = "regex")]
    Regex(String),
}

96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
mod token_serde {
    use super::*;
    use serde::de;
    use serde::Deserializer;
    use serde_json::Value;

    pub fn deserialize<'de, D>(deserializer: D) -> Result<Option<String>, D::Error>
    where
        D: Deserializer<'de>,
    {
        let value = Value::deserialize(deserializer)?;

        match value {
            Value::String(s) => Ok(Some(s)),
            Value::Object(map) => {
                if let Some(content) = map.get("content").and_then(|v| v.as_str()) {
                    Ok(Some(content.to_string()))
                } else {
                    Err(de::Error::custom(
                        "content key not found in structured token",
                    ))
                }
            }
            _ => Err(de::Error::custom("invalid token format")),
        }
    }
}

124
125
#[derive(Clone, Debug, Serialize, ToSchema)]
pub struct Info {
126
    /// Model info
127
128
129
130
    #[schema(example = "bigscience/blomm-560m")]
    pub model_id: String,
    #[schema(nullable = true, example = "e985a63cdc139290c5f700ff1929f0b5942cced2")]
    pub model_sha: Option<String>,
131
132
133
134
    #[schema(example = "torch.float16")]
    pub model_dtype: String,
    #[schema(example = "cuda")]
    pub model_device_type: String,
135
136
    #[schema(nullable = true, example = "text-generation")]
    pub model_pipeline_tag: Option<String>,
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
    /// Router Parameters
    #[schema(example = "128")]
    pub max_concurrent_requests: usize,
    #[schema(example = "2")]
    pub max_best_of: usize,
    #[schema(example = "4")]
    pub max_stop_sequences: usize,
    #[schema(example = "1024")]
    pub max_input_length: usize,
    #[schema(example = "2048")]
    pub max_total_tokens: usize,
    #[schema(example = "1.2")]
    pub waiting_served_ratio: f32,
    #[schema(example = "32000")]
    pub max_batch_total_tokens: u32,
    #[schema(example = "20")]
    pub max_waiting_tokens: usize,
154
155
    #[schema(nullable = true, example = "null")]
    pub max_batch_size: Option<usize>,
156
157
    #[schema(example = "2")]
    pub validation_workers: usize,
158
159
    #[schema(example = "32")]
    pub max_client_batch_size: usize,
160
    /// Router Info
161
162
163
164
    #[schema(example = "0.5.0")]
    pub version: &'static str,
    #[schema(nullable = true, example = "null")]
    pub sha: Option<&'static str>,
165
166
    #[schema(nullable = true, example = "null")]
    pub docker_label: Option<&'static str>,
167
168
}

drbh's avatar
drbh committed
169
#[derive(Clone, Debug, Deserialize, ToSchema, Default)]
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
170
pub(crate) struct GenerateParameters {
171
172
173
    #[serde(default)]
    #[schema(exclusive_minimum = 0, nullable = true, default = "null", example = 1)]
    pub best_of: Option<usize>,
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
    #[serde(default)]
    #[schema(
        exclusive_minimum = 0.0,
        nullable = true,
        default = "null",
        example = 0.5
    )]
    pub temperature: Option<f32>,
    #[serde(default)]
    #[schema(
        exclusive_minimum = 0.0,
        nullable = true,
        default = "null",
        example = 1.03
    )]
    pub repetition_penalty: Option<f32>,
    #[serde(default)]
191
192
193
194
195
196
197
198
    #[schema(
        exclusive_minimum = -2.0,
        nullable = true,
        default = "null",
        example = 0.1
    )]
    pub frequency_penalty: Option<f32>,
    #[serde(default)]
199
200
201
202
203
204
205
206
207
208
209
    #[schema(exclusive_minimum = 0, nullable = true, default = "null", example = 10)]
    pub top_k: Option<i32>,
    #[serde(default)]
    #[schema(
        exclusive_minimum = 0.0,
        maximum = 1.0,
        nullable = true,
        default = "null",
        example = 0.95
    )]
    pub top_p: Option<f32>,
210
    #[serde(default)]
211
212
213
214
215
216
217
218
219
    #[schema(
        exclusive_minimum = 0.0,
        maximum = 1.0,
        nullable = true,
        default = "null",
        example = 0.95
    )]
    pub typical_p: Option<f32>,
    #[serde(default)]
220
    #[schema(default = "false", example = true)]
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
221
222
    pub do_sample: bool,
    #[serde(default = "default_max_new_tokens")]
223
    #[schema(nullable = true, default = "100", example = "20")]
224
    pub max_new_tokens: Option<u32>,
OlivierDehaene's avatar
OlivierDehaene committed
225
    #[serde(default)]
226
    #[schema(nullable = true, default = "null", example = false)]
227
228
    pub return_full_text: Option<bool>,
    #[serde(default)]
229
    #[schema(inline, max_items = 4, example = json ! (["photographer"]))]
230
    pub stop: Vec<String>,
OlivierDehaene's avatar
OlivierDehaene committed
231
    #[serde(default)]
232
    #[schema(nullable = true, default = "null", example = "null")]
233
234
    pub truncate: Option<usize>,
    #[serde(default)]
235
236
237
    #[schema(default = "false", example = true)]
    pub watermark: bool,
    #[serde(default)]
238
    #[schema(default = "true")]
OlivierDehaene's avatar
OlivierDehaene committed
239
    pub details: bool,
240
    #[serde(default)]
241
242
243
    #[schema(default = "true")]
    pub decoder_input_details: bool,
    #[serde(default)]
244
245
246
247
248
249
    #[schema(
        exclusive_minimum = 0,
        nullable = true,
        default = "null",
        example = "null"
    )]
250
    pub seed: Option<u64>,
Nicolas Patry's avatar
Nicolas Patry committed
251
252
253
    #[serde(default)]
    #[schema(exclusive_minimum = 0, nullable = true, default = "null", example = 5)]
    pub top_n_tokens: Option<u32>,
drbh's avatar
drbh committed
254
255
    #[serde(default)]
    pub grammar: Option<GrammarType>,
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
256
257
}

258
fn default_max_new_tokens() -> Option<u32> {
259
    Some(100)
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
260
261
262
263
}

fn default_parameters() -> GenerateParameters {
    GenerateParameters {
264
        best_of: None,
265
266
        temperature: None,
        repetition_penalty: None,
267
        frequency_penalty: None,
268
269
        top_k: None,
        top_p: None,
270
        typical_p: None,
271
        do_sample: true,
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
272
        max_new_tokens: default_max_new_tokens(),
273
        return_full_text: None,
274
        stop: Vec::new(),
275
        truncate: None,
276
        watermark: false,
OlivierDehaene's avatar
OlivierDehaene committed
277
        details: false,
278
        decoder_input_details: false,
279
        seed: None,
Nicolas Patry's avatar
Nicolas Patry committed
280
        top_n_tokens: None,
drbh's avatar
drbh committed
281
        grammar: None,
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
282
283
284
    }
}

285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
mod prompt_serde {
    use serde::{self, Deserialize, Deserializer};
    use serde_json::Value;

    pub fn deserialize<'de, D>(deserializer: D) -> Result<Vec<String>, D::Error>
    where
        D: Deserializer<'de>,
    {
        let value = Value::deserialize(deserializer)?;
        match value {
            Value::String(s) => Ok(vec![s]),
            Value::Array(arr) if arr.is_empty() => Err(serde::de::Error::custom(
                "Empty array detected. Do not use an empty array for the prompt.",
            )),
            Value::Array(arr) => arr
                .iter()
                .map(|v| match v {
                    Value::String(s) => Ok(s.to_owned()),
                    _ => Err(serde::de::Error::custom("Expected a string")),
                })
                .collect(),
            _ => Err(serde::de::Error::custom(
                "Expected a string or an array of strings",
            )),
        }
    }
}

313
314
315
316
317
318
319
320
321
#[derive(Clone, Deserialize, Serialize, ToSchema, Debug)]
pub struct CompletionRequest {
    /// UNUSED
    #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
    /// ID of the model to use. See the model endpoint compatibility table for details on which models work with the Chat API.
    pub model: String,

    /// The prompt to generate completions for.
    #[schema(example = "What is Deep Learning?")]
322
323
    #[serde(deserialize_with = "prompt_serde::deserialize")]
    pub prompt: Vec<String>,
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383

    /// The maximum number of tokens that can be generated in the chat completion.
    #[serde(default)]
    #[schema(default = "32")]
    pub max_tokens: Option<u32>,

    /// What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while
    /// lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or `top_p` but not both.
    #[serde(default)]
    #[schema(nullable = true, example = 1.0)]
    pub temperature: Option<f32>,

    /// An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the
    /// tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.
    #[serde(default)]
    #[schema(nullable = true, example = 0.95)]
    pub top_p: Option<f32>,

    #[serde(default = "bool::default")]
    pub stream: bool,

    #[schema(nullable = true, example = 42)]
    pub seed: Option<u64>,

    /// The text to append to the prompt. This is useful for completing sentences or generating a paragraph of text.
    /// please see the completion_template field in the model's tokenizer_config.json file for completion template.
    #[serde(default)]
    pub suffix: Option<String>,

    #[serde(default)]
    pub repetition_penalty: Option<f32>,

    /// Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far,
    /// decreasing the model's likelihood to repeat the same line verbatim.
    #[serde(default)]
    #[schema(example = "1.0")]
    pub frequency_penalty: Option<f32>,
}

#[derive(Clone, Deserialize, Serialize, ToSchema, Default)]
pub(crate) struct Completion {
    pub id: String,
    pub object: String,
    #[schema(example = "1706270835")]
    pub created: u64,
    #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
    pub model: String,
    pub system_fingerprint: String,
    pub choices: Vec<CompletionComplete>,
    pub usage: Usage,
}

#[derive(Clone, Deserialize, Serialize, ToSchema)]
pub(crate) struct CompletionComplete {
    pub index: u32,
    pub text: String,
    pub logprobs: Option<Vec<f32>>,
    pub finish_reason: String,
}

384
#[derive(Clone, Deserialize, Serialize, ToSchema)]
385
386
387
pub(crate) struct ChatCompletion {
    pub id: String,
    pub object: String,
388
    #[schema(example = "1706270835")]
389
    pub created: u64,
390
    #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
391
392
393
394
395
396
    pub model: String,
    pub system_fingerprint: String,
    pub choices: Vec<ChatCompletionComplete>,
    pub usage: Usage,
}

397
#[derive(Clone, Deserialize, Serialize, ToSchema)]
398
399
400
pub(crate) struct ChatCompletionComplete {
    pub index: u32,
    pub message: Message,
401
    pub logprobs: Option<ChatCompletionLogprobs>,
402
403
404
    pub finish_reason: String,
}

405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
#[derive(Clone, Deserialize, Serialize, ToSchema)]
pub(crate) struct ChatCompletionLogprobs {
    content: Vec<ChatCompletionLogprob>,
}

impl From<(Token, Vec<Token>)> for ChatCompletionLogprobs {
    fn from(value: (Token, Vec<Token>)) -> Self {
        let (token, top_tokens) = value;

        Self {
            content: vec![ChatCompletionLogprob {
                token: token.text,
                logprob: token.logprob,
                top_logprobs: top_tokens
                    .into_iter()
                    .map(|t| ChatCompletionTopLogprob {
                        token: t.text,
                        logprob: t.logprob,
                    })
                    .collect(),
            }],
        }
    }
}

impl From<(Vec<Token>, Vec<Vec<Token>>)> for ChatCompletionLogprobs {
    fn from(value: (Vec<Token>, Vec<Vec<Token>>)) -> Self {
        let (tokens, top_tokens) = value;
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447

        // Create an iterator that produces None for top_tokens once it's exhausted
        let top_tokens_iter = top_tokens
            .into_iter()
            .map(Some)
            .chain(std::iter::repeat(None));

        let content = tokens
            .into_iter()
            .zip(top_tokens_iter)
            .map(|(t, top_t_option)| ChatCompletionLogprob {
                token: t.text,
                logprob: t.logprob,
                top_logprobs: match top_t_option {
                    Some(top_t) => top_t
448
449
450
451
452
453
                        .into_iter()
                        .map(|t| ChatCompletionTopLogprob {
                            token: t.text,
                            logprob: t.logprob,
                        })
                        .collect(),
454
455
456
457
458
459
                    None => vec![], // Handle the case where there are no top tokens
                },
            })
            .collect();

        Self { content }
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
    }
}

#[derive(Clone, Deserialize, Serialize, ToSchema)]
pub(crate) struct ChatCompletionLogprob {
    token: String,
    logprob: f32,
    top_logprobs: Vec<ChatCompletionTopLogprob>,
}

#[derive(Clone, Deserialize, Serialize, ToSchema)]
pub(crate) struct ChatCompletionTopLogprob {
    token: String,
    logprob: f32,
}

476
#[derive(Clone, Deserialize, Serialize, ToSchema, Default)]
477
478
479
480
481
482
483
484
485
486
pub(crate) struct Usage {
    pub prompt_tokens: u32,
    pub completion_tokens: u32,
    pub total_tokens: u32,
}

impl ChatCompletion {
    pub(crate) fn new(
        model: String,
        system_fingerprint: String,
drbh's avatar
drbh committed
487
        output: Option<String>,
488
489
490
        created: u64,
        details: Details,
        return_logprobs: bool,
491
        tool_calls: Option<Vec<ToolCall>>,
492
493
494
495
496
497
498
499
500
501
502
503
    ) -> Self {
        Self {
            id: String::new(),
            object: "text_completion".into(),
            created,
            model,
            system_fingerprint,
            choices: vec![ChatCompletionComplete {
                index: 0,
                message: Message {
                    role: "assistant".into(),
                    content: output,
504
                    name: None,
drbh's avatar
drbh committed
505
                    tool_calls,
506
507
                },
                logprobs: return_logprobs
508
                    .then(|| ChatCompletionLogprobs::from((details.tokens, details.top_tokens))),
509
510
511
512
513
514
515
516
517
518
                finish_reason: details.finish_reason.to_string(),
            }],
            usage: Usage {
                prompt_tokens: details.prefill.len() as u32,
                completion_tokens: details.generated_tokens,
                total_tokens: details.prefill.len() as u32 + details.generated_tokens,
            },
        }
    }
}
519
520
521
522
523
524
525
526
527
#[derive(Clone, Deserialize, Serialize, ToSchema)]
pub(crate) struct CompletionCompleteChunk {
    pub id: String,
    pub object: String,
    pub created: u64,
    pub choices: Vec<CompletionComplete>,
    pub model: String,
    pub system_fingerprint: String,
}
528
#[derive(Clone, Deserialize, Serialize, ToSchema)]
529
530
531
pub(crate) struct ChatCompletionChunk {
    pub id: String,
    pub object: String,
532
    #[schema(example = "1706270978")]
533
    pub created: u64,
534
    #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
535
536
537
538
539
    pub model: String,
    pub system_fingerprint: String,
    pub choices: Vec<ChatCompletionChoice>,
}

540
#[derive(Clone, Deserialize, Serialize, ToSchema)]
541
542
543
pub(crate) struct ChatCompletionChoice {
    pub index: u32,
    pub delta: ChatCompletionDelta,
544
    pub logprobs: Option<ChatCompletionLogprobs>,
545
546
547
    pub finish_reason: Option<String>,
}

548
#[derive(Clone, Debug, Deserialize, Serialize, ToSchema)]
549
pub(crate) struct ChatCompletionDelta {
550
    #[schema(example = "user")]
551
    pub role: String,
drbh's avatar
drbh committed
552
    #[serde(default, skip_serializing_if = "Option::is_none")]
553
    #[schema(example = "What is Deep Learning?")]
drbh's avatar
drbh committed
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
    pub content: Option<String>,
    // default to None
    #[serde(default, skip_serializing_if = "Option::is_none")]
    pub tool_calls: Option<DeltaToolCall>,
}

#[derive(Clone, Deserialize, Serialize, ToSchema, Debug)]
pub(crate) struct DeltaToolCall {
    pub index: u32,
    pub id: String,
    pub r#type: String,
    pub function: Function,
}

#[derive(Clone, Deserialize, Serialize, ToSchema, Debug)]
pub(crate) struct Function {
    pub name: Option<String>,
    pub arguments: String,
572
573
}

drbh's avatar
drbh committed
574
#[allow(clippy::too_many_arguments)]
575
576
577
578
impl ChatCompletionChunk {
    pub(crate) fn new(
        model: String,
        system_fingerprint: String,
drbh's avatar
drbh committed
579
580
        delta: Option<String>,
        tool_calls: Option<Vec<String>>,
581
        created: u64,
582
        logprobs: Option<ChatCompletionLogprobs>,
583
584
585
586
587
588
589
590
591
        finish_reason: Option<String>,
    ) -> Self {
        Self {
            id: String::new(),
            object: "text_completion".to_string(),
            created,
            model,
            system_fingerprint,
            choices: vec![ChatCompletionChoice {
592
                index: 0,
593
594
595
                delta: ChatCompletionDelta {
                    role: "assistant".to_string(),
                    content: delta,
drbh's avatar
drbh committed
596
                    tool_calls: tool_calls.map(|tc| DeltaToolCall {
597
                        index: 0,
drbh's avatar
drbh committed
598
599
600
601
602
603
604
                        id: String::new(),
                        r#type: "function".to_string(),
                        function: Function {
                            name: None,
                            arguments: tc[0].to_string(),
                        },
                    }),
605
606
607
608
609
610
611
612
613
614
                },
                logprobs,
                finish_reason,
            }],
        }
    }
}

#[derive(Clone, Deserialize, ToSchema, Serialize)]
pub(crate) struct ChatRequest {
615
    #[schema(example = "mistralai/Mistral-7B-Instruct-v0.2")]
drbh's avatar
drbh committed
616
    /// [UNUSED] ID of the model to use. See the model endpoint compatibility table for details on which models work with the Chat API.
617
    pub model: String,
drbh's avatar
drbh committed
618

619
    /// A list of messages comprising the conversation so far.
drbh's avatar
drbh committed
620
    #[schema(example = "[{\"role\": \"user\", \"content\": \"What is Deep Learning?\"}]")]
621
622
623
624
625
    pub messages: Vec<Message>,

    /// Number between -2.0 and 2.0. Positive values penalize new tokens based on their existing frequency in the text so far,
    /// decreasing the model's likelihood to repeat the same line verbatim.
    #[serde(default)]
626
    #[schema(example = "1.0")]
627
628
629
630
631
632
633
634
635
636
637
638
639
640
    pub frequency_penalty: Option<f32>,

    /// UNUSED
    /// Modify the likelihood of specified tokens appearing in the completion. Accepts a JSON object that maps tokens
    /// (specified by their token ID in the tokenizer) to an associated bias value from -100 to 100. Mathematically,
    /// the bias is added to the logits generated by the model prior to sampling. The exact effect will vary per model,
    /// but values between -1 and 1 should decrease or increase likelihood of selection; values like -100 or 100 should
    /// result in a ban or exclusive selection of the relevant token.
    #[serde(default)]
    pub logit_bias: Option<Vec<f32>>,

    /// Whether to return log probabilities of the output tokens or not. If true, returns the log probabilities of each
    /// output token returned in the content of message.
    #[serde(default)]
641
    #[schema(example = "false")]
642
643
644
645
646
    pub logprobs: Option<bool>,

    /// An integer between 0 and 5 specifying the number of most likely tokens to return at each token position, each with
    /// an associated log probability. logprobs must be set to true if this parameter is used.
    #[serde(default)]
647
    #[schema(example = "5")]
648
649
650
651
    pub top_logprobs: Option<u32>,

    /// The maximum number of tokens that can be generated in the chat completion.
    #[serde(default)]
652
    #[schema(example = "32")]
653
654
655
656
657
658
    pub max_tokens: Option<u32>,

    /// UNUSED
    /// How many chat completion choices to generate for each input message. Note that you will be charged based on the
    /// number of generated tokens across all of the choices. Keep n as 1 to minimize costs.
    #[serde(default)]
659
    #[schema(nullable = true, example = "2")]
660
661
662
663
664
    pub n: Option<u32>,

    /// Number between -2.0 and 2.0. Positive values penalize new tokens based on whether they appear in the text so far,
    /// increasing the model's likelihood to talk about new topics
    #[serde(default)]
665
    #[schema(nullable = true, example = 0.1)]
666
667
    pub presence_penalty: Option<f32>,

668
669
670
671
672
    /// Up to 4 sequences where the API will stop generating further tokens.
    #[serde(default)]
    #[schema(nullable = true, example = "null")]
    pub stop: Option<Vec<String>>,

673
674
675
676
677
    #[serde(default = "bool::default")]
    pub stream: bool,

    #[schema(nullable = true, example = 42)]
    pub seed: Option<u64>,
678
679
680
681
682
683

    /// What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while
    /// lower values like 0.2 will make it more focused and deterministic.
    ///
    /// We generally recommend altering this or `top_p` but not both.
    #[serde(default)]
684
    #[schema(nullable = true, example = 1.0)]
685
686
687
688
689
    pub temperature: Option<f32>,

    /// An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the
    /// tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.
    #[serde(default)]
690
    #[schema(nullable = true, example = 0.95)]
691
    pub top_p: Option<f32>,
drbh's avatar
drbh committed
692
693
694
695
696
697
698
699
700
701
702

    /// A list of tools the model may call. Currently, only functions are supported as a tool. Use this to provide a list of
    /// functions the model may generate JSON inputs for.
    #[serde(default)]
    #[schema(nullable = true, example = "null")]
    pub tools: Option<Vec<Tool>>,

    /// A prompt to be appended before the tools
    #[serde(default = "default_tool_prompt")]
    #[schema(
        nullable = true,
703
        example = "\"You will be presented with a JSON schema representing a set of tools.\nIf the user request lacks of sufficient information to make a precise tool selection: Do not invent any tool's properties, instead notify with an error message.\n\nJSON Schema:\n\""
drbh's avatar
drbh committed
704
705
706
707
708
709
710
711
712
713
714
715
    )]
    pub tool_prompt: Option<String>,

    /// A specific tool to use. If not provided, the model will default to use any of the tools provided in the tools parameter.
    #[serde(default)]
    #[schema(nullable = true, example = "null")]
    #[serde(deserialize_with = "deserialize_tool_choice::deserialize")]
    pub tool_choice: Option<ToolType>,
}

fn default_tool_prompt() -> Option<String> {
    Some(
716
        "\nYou will be presented with a JSON schema representing a set of tools.\nIf the user request lacks of sufficient information to make a precise tool selection: Do not invent any tool's properties, instead notify with an error message.\n\nJSON Schema:\n".to_string(),
drbh's avatar
drbh committed
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
    )
}
#[derive(Clone, Deserialize, ToSchema, Serialize)]
enum ToolType {
    FunctionName(String),
    OneOf,
}

/// Deserialize the tool choice from the JSON input or from the function name ("none" is allowed but mapped to None)
mod deserialize_tool_choice {
    use super::*;
    use serde::de;
    use serde::Deserializer;
    use serde_json::Value;

    pub fn deserialize<'de, D>(deserializer: D) -> Result<Option<ToolType>, D::Error>
    where
        D: Deserializer<'de>,
    {
        let value = Value::deserialize(deserializer)?;

        match value {
            Value::String(s) => match s.as_str() {
                "none" => Ok(None),
                "auto" => Ok(Some(ToolType::OneOf)),
                _ => Ok(Some(ToolType::FunctionName(s))),
            },
            Value::Object(map) => {
                if let Some(content) = map
                    .get("function")
                    .and_then(|v| v.get("name"))
                    .and_then(|v| v.as_str())
                {
                    Ok(Some(ToolType::FunctionName(content.to_string())))
                } else {
                    Err(de::Error::custom("function key not found in tool choice"))
                }
            }
            Value::Null => Ok(Some(ToolType::OneOf)),
            _ => Err(de::Error::custom("invalid token format")),
        }
    }
}

761
#[derive(Debug, Deserialize, Serialize, ToSchema, PartialEq)]
drbh's avatar
drbh committed
762
763
764
765
766
767
pub struct Tools {
    #[serde(flatten)]
    functions_map: FunctionsMap,
    properties: Properties,
}

768
#[derive(Debug, Serialize, Deserialize, PartialEq)]
drbh's avatar
drbh committed
769
770
771
772
773
struct FunctionsMap {
    #[serde(rename = "$functions")]
    functions: std::collections::HashMap<String, serde_json::Value>,
}

774
#[derive(Debug, Serialize, Deserialize, PartialEq)]
drbh's avatar
drbh committed
775
776
777
778
779
struct FunctionRef {
    #[serde(rename = "$ref")]
    ref_path: String,
}

780
#[derive(Debug, Serialize, Deserialize, PartialEq)]
drbh's avatar
drbh committed
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
struct Properties {
    #[serde(serialize_with = "serialize_function")]
    function: Vec<FunctionRef>,
}

fn serialize_function<S>(functions: &Vec<FunctionRef>, serializer: S) -> Result<S::Ok, S::Error>
where
    S: serde::Serializer,
{
    use serde::ser::SerializeStruct;
    let mut state = serializer.serialize_struct("Function", 1)?;
    state.serialize_field("anyOf", functions)?;
    state.end()
}

#[derive(Clone, Debug, Deserialize, Serialize, ToSchema, Default)]
pub(crate) struct FunctionDefinition {
    #[serde(default)]
    pub description: Option<String>,
    pub name: String,
801
802
    #[serde(alias = "parameters")]
    pub arguments: serde_json::Value,
drbh's avatar
drbh committed
803
804
805
806
807
808
809
810
811
}

#[derive(Clone, Debug, Deserialize, Serialize, ToSchema)]
pub(crate) struct Tool {
    // The type of the tool. Currently, only 'function' is supported.
    #[schema(example = "function")]
    pub r#type: String,
    // Grab the tool as generic JSON for debugging purposes.
    pub function: FunctionDefinition,
812
813
}

814
#[derive(Clone, Serialize, Deserialize, Default)]
815
816
817
818
pub(crate) struct ChatTemplateInputs<'a> {
    messages: Vec<Message>,
    bos_token: Option<&'a str>,
    eos_token: Option<&'a str>,
819
    add_generation_prompt: bool,
820
821
    tools: Option<&'a str>,
    tools_prompt: Option<&'a str>,
822
823
}

drbh's avatar
drbh committed
824
825
826
827
828
829
830
#[derive(Clone, Deserialize, Serialize, ToSchema, Default, Debug)]
pub(crate) struct ToolCall {
    pub id: u32,
    pub r#type: String,
    pub function: FunctionDefinition,
}

831
832
833
834
#[derive(Clone, Deserialize, ToSchema, Serialize)]
pub(crate) struct Message {
    #[schema(example = "user")]
    pub role: String,
drbh's avatar
drbh committed
835
    #[serde(skip_serializing_if = "Option::is_none")]
836
    #[schema(example = "My name is David and I")]
drbh's avatar
drbh committed
837
    pub content: Option<String>,
drbh's avatar
drbh committed
838
    #[serde(default, skip_serializing_if = "Option::is_none")]
839
840
    #[schema(example = "\"David\"")]
    pub name: Option<String>,
drbh's avatar
drbh committed
841
    #[serde(default, skip_serializing_if = "Option::is_none")]
842
    pub tool_calls: Option<Vec<ToolCall>>,
843
844
}

845
#[derive(Clone, Debug, Deserialize, ToSchema)]
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
846
pub(crate) struct GenerateRequest {
847
    #[schema(example = "My name is Olivier and I")]
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
848
849
850
851
852
    pub inputs: String,
    #[serde(default = "default_parameters")]
    pub parameters: GenerateParameters,
}

853
854
855
856
857
858
859
#[derive(Clone, Debug, Deserialize, ToSchema)]
pub(crate) struct CompatGenerateRequest {
    #[schema(example = "My name is Olivier and I")]
    pub inputs: String,
    #[serde(default = "default_parameters")]
    pub parameters: GenerateParameters,
    #[serde(default)]
OlivierDehaene's avatar
OlivierDehaene committed
860
    #[schema(default = "false")]
861
862
863
864
865
866
867
868
869
870
871
872
    pub stream: bool,
}

impl From<CompatGenerateRequest> for GenerateRequest {
    fn from(req: CompatGenerateRequest) -> Self {
        Self {
            inputs: req.inputs,
            parameters: req.parameters,
        }
    }
}

873
874
875
876
877
878
#[derive(Debug, Serialize, ToSchema)]
pub struct PrefillToken {
    #[schema(example = 0)]
    id: u32,
    #[schema(example = "test")]
    text: String,
879
    #[schema(nullable = true, example = - 0.34)]
880
881
882
    logprob: f32,
}

883
#[derive(Debug, Serialize, ToSchema, Clone)]
884
885
886
887
888
pub struct Token {
    #[schema(example = 0)]
    id: u32,
    #[schema(example = "test")]
    text: String,
889
    #[schema(nullable = true, example = - 0.34)]
890
    logprob: f32,
891
892
    #[schema(example = "false")]
    special: bool,
893
894
}

895
896
897
898
899
900
901
902
903
904
905
906
#[derive(Debug, Serialize, ToSchema)]
pub struct SimpleToken {
    #[schema(example = 0)]
    id: u32,
    #[schema(example = "test")]
    text: String,
    #[schema(example = 0)]
    start: usize,
    #[schema(example = 2)]
    stop: usize,
}

907
908
#[derive(Serialize, ToSchema)]
#[serde(rename_all(serialize = "snake_case"))]
909
#[schema(example = "Length")]
910
911
912
913
914
915
916
917
918
pub(crate) enum FinishReason {
    #[schema(rename = "length")]
    Length,
    #[serde(rename = "eos_token")]
    #[schema(rename = "eos_token")]
    EndOfSequenceToken,
    #[schema(rename = "stop_sequence")]
    StopSequence,
}
919

920
921
922
923
924
925
926
927
928
929
impl std::fmt::Display for FinishReason {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match self {
            FinishReason::Length => write!(f, "length"),
            FinishReason::EndOfSequenceToken => write!(f, "eos_token"),
            FinishReason::StopSequence => write!(f, "stop_sequence"),
        }
    }
}

930
931
932
933
934
935
936
937
938
939
940
941
#[derive(Serialize, ToSchema)]
pub(crate) struct BestOfSequence {
    #[schema(example = "test")]
    pub generated_text: String,
    #[schema(example = "length")]
    pub finish_reason: FinishReason,
    #[schema(example = 1)]
    pub generated_tokens: u32,
    #[schema(nullable = true, example = 42)]
    pub seed: Option<u64>,
    pub prefill: Vec<PrefillToken>,
    pub tokens: Vec<Token>,
Nicolas Patry's avatar
Nicolas Patry committed
942
943
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub top_tokens: Vec<Vec<Token>>,
944
945
}

946
#[derive(Serialize, ToSchema)]
OlivierDehaene's avatar
OlivierDehaene committed
947
pub(crate) struct Details {
948
949
950
    #[schema(example = "length")]
    pub finish_reason: FinishReason,
    #[schema(example = 1)]
OlivierDehaene's avatar
OlivierDehaene committed
951
    pub generated_tokens: u32,
952
    #[schema(nullable = true, example = 42)]
953
    pub seed: Option<u64>,
954
955
    pub prefill: Vec<PrefillToken>,
    pub tokens: Vec<Token>,
956
957
    #[serde(skip_serializing_if = "Option::is_none")]
    pub best_of_sequences: Option<Vec<BestOfSequence>>,
Nicolas Patry's avatar
Nicolas Patry committed
958
959
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub top_tokens: Vec<Vec<Token>>,
OlivierDehaene's avatar
OlivierDehaene committed
960
961
}

962
#[derive(Serialize, ToSchema)]
963
pub(crate) struct GenerateResponse {
964
    #[schema(example = "test")]
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
965
    pub generated_text: String,
OlivierDehaene's avatar
OlivierDehaene committed
966
967
    #[serde(skip_serializing_if = "Option::is_none")]
    pub details: Option<Details>,
Olivier Dehaene's avatar
v0.1.0  
Olivier Dehaene committed
968
}
969

970
971
972
973
#[derive(Serialize, ToSchema)]
#[serde(transparent)]
pub(crate) struct TokenizeResponse(Vec<SimpleToken>);

974
975
976
977
978
979
#[derive(Serialize, ToSchema)]
pub(crate) struct StreamDetails {
    #[schema(example = "length")]
    pub finish_reason: FinishReason,
    #[schema(example = 1)]
    pub generated_tokens: u32,
980
    #[schema(nullable = true, example = 42)]
981
982
983
984
    pub seed: Option<u64>,
}

#[derive(Serialize, ToSchema)]
985
pub(crate) struct StreamResponse {
986
    pub index: u32,
987
    pub token: Token,
Nicolas Patry's avatar
Nicolas Patry committed
988
989
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub top_tokens: Vec<Token>,
990
    #[schema(nullable = true, default = "null", example = "test")]
991
    pub generated_text: Option<String>,
992
993
    #[schema(nullable = true, default = "null")]
    pub details: Option<StreamDetails>,
994
995
}

996
#[derive(Serialize, ToSchema)]
997
998
pub(crate) struct ErrorResponse {
    pub error: String,
999
    pub error_type: String,
1000
}
1001
1002

#[cfg(test)]
1003
mod tests {
1004
1005
    use super::*;

1006
1007
    use tokenizers::Tokenizer;

1008
    pub(crate) async fn get_tokenizer() -> Tokenizer {
1009
1010
1011
1012
        let api = hf_hub::api::sync::Api::new().unwrap();
        let repo = api.model("gpt2".to_string());
        let filename = repo.get("tokenizer.json").unwrap();
        Tokenizer::from_file(filename).unwrap()
1013
    }
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027

    #[test]
    fn test_hub_nested_tokens_tokenizer_config() {
        // this is a subset of the tokenizer.json file
        // in this case we expect the tokens to be encoded as simple strings
        let json_content = r#"{
            "chat_template": "test",
            "bos_token": "<|begin▁of▁sentence|>",
            "eos_token": "<|end▁of▁sentence|>"
        }"#;

        let config: HubTokenizerConfig = serde_json::from_str(json_content).unwrap();

        // check that we successfully parsed the tokens
1028
1029
1030
1031
        assert_eq!(
            config.chat_template,
            Some(ChatTemplateVersions::Single("test".to_string()))
        );
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
        assert_eq!(
            config.bos_token,
            Some("<|begin▁of▁sentence|>".to_string())
        );
        assert_eq!(config.eos_token, Some("<|end▁of▁sentence|>".to_string()));

        // in this case we expect the tokens to be encoded as structured tokens
        // we want the content of the structured token
        let json_content = r#"{
            "chat_template": "test",
            "bos_token": {
              "__type": "AddedToken",
              "content": "<|begin▁of▁sentence|>",
              "lstrip": false,
              "normalized": true,
              "rstrip": false,
              "single_word": false
            },
            "eos_token": {
              "__type": "AddedToken",
              "content": "<|end▁of▁sentence|>",
              "lstrip": false,
              "normalized": true,
              "rstrip": false,
              "single_word": false
            }
        }"#;

        let config: HubTokenizerConfig = serde_json::from_str(json_content).unwrap();

        // check that we successfully parsed the tokens
1063
1064
1065
1066
        assert_eq!(
            config.chat_template,
            Some(ChatTemplateVersions::Single("test".to_string()))
        );
1067
1068
1069
1070
1071
1072
        assert_eq!(
            config.bos_token,
            Some("<|begin▁of▁sentence|>".to_string())
        );
        assert_eq!(config.eos_token, Some("<|end▁of▁sentence|>".to_string()));
    }
1073
}