API

llama.rn

llama.rn

Table of contents

Classes

• LlamaContext
• SchemaGrammarConverter
• SchemaGrammarConverterBuiltinRule

Interfaces

• SchemaGrammarConverterPropOrder

Type Aliases

• BenchResult
• CompletionParams
• ContextParams
• EmbeddingParams
• NativeCompletionParams
• NativeCompletionResult
• NativeCompletionResultTimings
• NativeCompletionTokenProb
• NativeCompletionTokenProbItem
• NativeContextParams
• NativeEmbeddingParams
• NativeEmbeddingResult
• NativeLlamaContext
• NativeSessionLoadResult
• NativeTokenizeResult
• RNLlamaMessagePart
• RNLlamaOAICompatibleMessage
• TokenData

Functions

• convertJsonSchemaToGrammar
• initLlama
• loadLlamaModelInfo
• releaseAllLlama
• setContextLimit

Type Aliases

BenchResult

Ƭ BenchResult: Object

Type declaration

Name	Type
modelDesc	string
modelNParams	number
modelSize	number
ppAvg	number
ppStd	number
tgAvg	number
tgStd	number

Defined in

index.ts:84

---

CompletionParams

Ƭ CompletionParams: Omit<NativeCompletionParams, "emit_partial_completion" | "prompt"> & { chatTemplate?: string ; messages?: RNLlamaOAICompatibleMessage[] ; prompt?: string }

Defined in

index.ts:75

---

ContextParams

Ƭ ContextParams: Omit<NativeContextParams, "cache_type_k" | "cache_type_v" | "pooling_type"> & { cache_type_k?: "f16" | "f32" | "q8_0" | "q4_0" | "q4_1" | "iq4_nl" | "q5_0" | "q5_1" ; cache_type_v?: "f16" | "f32" | "q8_0" | "q4_0" | "q4_1" | "iq4_nl" | "q5_0" | "q5_1" ; pooling_type?: "none" | "mean" | "cls" | "last" | "rank" }

Defined in

index.ts:64

---

EmbeddingParams

Ƭ EmbeddingParams: NativeEmbeddingParams

Defined in

index.ts:73

---

NativeCompletionParams

Ƭ NativeCompletionParams: Object

Type declaration

Name	Type	Description
dry_allowed_length?	number	Tokens that extend repetition beyond this receive exponentially increasing penalty: multiplier * base ^ (length of repeating sequence before token - allowed length). Default: 2
dry_base?	number	Set the DRY repetition penalty base value. Default: 1.75
dry_multiplier?	number	Set the DRY (Don't Repeat Yourself) repetition penalty multiplier. Default: 0.0, which is disabled.
dry_penalty_last_n?	number	How many tokens to scan for repetitions. Default: -1, where 0 is disabled and -1 is context size.
dry_sequence_breakers?	string[]	Specify an array of sequence breakers for DRY sampling. Only a JSON array of strings is accepted. Default: ['\n', ':', '"', '*']
emit_partial_completion	boolean	-
grammar?	string	Set grammar for grammar-based sampling. Default: no grammar
ignore_eos?	boolean	Ignore end of stream token and continue generating. Default: false
logit_bias?	number[][]	Modify the likelihood of a token appearing in the generated text completion. For example, use "logit_bias": [[15043,1.0]] to increase the likelihood of the token 'Hello', or "logit_bias": [[15043,-1.0]] to decrease its likelihood. Setting the value to false, "logit_bias": [[15043,false]] ensures that the token Hello is never produced. The tokens can also be represented as strings, e.g.[["Hello, World!",-0.5]] will reduce the likelihood of all the individual tokens that represent the string Hello, World!, just like the presence_penalty does. Default: []
min_p?	number	The minimum probability for a token to be considered, relative to the probability of the most likely token. Default: 0.05
mirostat?	number	Enable Mirostat sampling, controlling perplexity during text generation. Default: 0, where 0 is disabled, 1 is Mirostat, and 2 is Mirostat 2.0.
mirostat_eta?	number	Set the Mirostat learning rate, parameter eta. Default: 0.1
mirostat_tau?	number	Set the Mirostat target entropy, parameter tau. Default: 5.0
n_predict?	number	Set the maximum number of tokens to predict when generating text. Note: May exceed the set limit slightly if the last token is a partial multibyte character. When 0,no tokens will be generated but the prompt is evaluated into the cache. Default: -1, where -1 is infinity.
n_probs?	number	If greater than 0, the response also contains the probabilities of top N tokens for each generated token given the sampling settings. Note that for temperature < 0 the tokens are sampled greedily but token probabilities are still being calculated via a simple softmax of the logits without considering any other sampler settings. Default: 0
n_threads?	number	-
penalize_nl?	boolean	Penalize newline tokens when applying the repeat penalty. Default: false
penalty_freq?	number	Repeat alpha frequency penalty. Default: 0.0, which is disabled.
penalty_last_n?	number	Last n tokens to consider for penalizing repetition. Default: 64, where 0 is disabled and -1 is ctx-size.
penalty_present?	number	Repeat alpha presence penalty. Default: 0.0, which is disabled.
penalty_repeat?	number	Control the repetition of token sequences in the generated text. Default: 1.0
prompt	string	-
seed?	number	Set the random number generator (RNG) seed. Default: -1, which is a random seed.
stop?	string[]	Specify a JSON array of stopping strings. These words will not be included in the completion, so make sure to add them to the prompt for the next iteration. Default: []
temperature?	number	Adjust the randomness of the generated text. Default: 0.8
top_k?	number	Limit the next token selection to the K most probable tokens. Default: 40
top_p?	number	Limit the next token selection to a subset of tokens with a cumulative probability above a threshold P. Default: 0.95
typical_p?	number	Enable locally typical sampling with parameter p. Default: 1.0, which is disabled.
xtc_probability?	number	Set the chance for token removal via XTC sampler. Default: 0.0, which is disabled.
xtc_threshold?	number	Set a minimum probability threshold for tokens to be removed via XTC sampler. Default: 0.1 (> 0.5 disables XTC)

Defined in

NativeRNLlama.ts:50

---

NativeCompletionResult

Ƭ NativeCompletionResult: Object

Type declaration

Name	Type
completion_probabilities?	NativeCompletionTokenProb[]
stopped_eos	boolean
stopped_limit	number
stopped_word	string
stopping_word	string
text	string
timings	NativeCompletionResultTimings
tokens_cached	number
tokens_evaluated	number
tokens_predicted	number
truncated	boolean

Defined in

NativeRNLlama.ts:195

---

NativeCompletionResultTimings

Ƭ NativeCompletionResultTimings: Object

Type declaration

Name	Type
predicted_ms	number
predicted_n	number
predicted_per_second	number
predicted_per_token_ms	number
prompt_ms	number
prompt_n	number
prompt_per_second	number
prompt_per_token_ms	number

Defined in

NativeRNLlama.ts:184

---

NativeCompletionTokenProb

Ƭ NativeCompletionTokenProb: Object

Type declaration

Name	Type
content	string
probs	NativeCompletionTokenProbItem[]

Defined in

NativeRNLlama.ts:179

---

NativeCompletionTokenProbItem

Ƭ NativeCompletionTokenProbItem: Object

Type declaration

Name	Type
prob	number
tok_str	string

Defined in

NativeRNLlama.ts:174

---

NativeContextParams

Ƭ NativeContextParams: Object

Type declaration

Name	Type	Description
cache_type_k?	string	KV cache data type for the K (Experimental in llama.cpp)
cache_type_v?	string	KV cache data type for the V (Experimental in llama.cpp)
embd_normalize?	number	-
embedding?	boolean	-
flash_attn?	boolean	Enable flash attention, only recommended in GPU device (Experimental in llama.cpp)
is_model_asset?	boolean	-
lora?	string	-
lora_scaled?	number	-
model	string	-
n_batch?	number	-
n_ctx?	number	-
n_gpu_layers?	number	-
n_threads?	number	-
pooling_type?	number	-
rope_freq_base?	number	-
rope_freq_scale?	number	-
use_mlock?	boolean	-
use_mmap?	boolean	-
use_progress_callback?	boolean	-
vocab_only?	boolean	-

Defined in

NativeRNLlama.ts:8

---

NativeEmbeddingParams

Ƭ NativeEmbeddingParams: Object

Type declaration

Name	Type
embd_normalize?	number

Defined in

NativeRNLlama.ts:4

---

NativeEmbeddingResult

Ƭ NativeEmbeddingResult: Object

Type declaration

Name	Type
embedding	number[]

Defined in

NativeRNLlama.ts:215

---

NativeLlamaContext

Ƭ NativeLlamaContext: Object

Type declaration

Name	Type
contextId	number
gpu	boolean
model	Object
reasonNoGPU	string

Defined in

NativeRNLlama.ts:219

---

NativeSessionLoadResult

Ƭ NativeSessionLoadResult: Object

Type declaration

Name	Type
prompt	string
tokens_loaded	number

Defined in

NativeRNLlama.ts:226

---

NativeTokenizeResult

Ƭ NativeTokenizeResult: Object

Type declaration

Name	Type
tokens	number[]

Defined in

NativeRNLlama.ts:211

---

RNLlamaMessagePart

Ƭ RNLlamaMessagePart: Object

Type declaration

Name	Type
text?	string

Defined in

chat.ts:3

---

RNLlamaOAICompatibleMessage

Ƭ RNLlamaOAICompatibleMessage: Object

Type declaration

Name	Type
content?	string | RNLlamaMessagePart[] | any
role	string

Defined in

chat.ts:7

---

TokenData

Ƭ TokenData: Object

Type declaration

Name	Type
completion_probabilities?	NativeCompletionTokenProb[]
token	string

Defined in

index.ts:54

Functions

convertJsonSchemaToGrammar

▸ convertJsonSchemaToGrammar(«destructured»): string | Promise<string>

Parameters

Name	Type
«destructured»	Object
› allowFetch?	boolean
› dotall?	boolean
› propOrder?	SchemaGrammarConverterPropOrder
› schema	any

Returns

string | Promise<string>

Defined in

grammar.ts:826

---

initLlama

▸ initLlama(«destructured», onProgress?): Promise<LlamaContext>

Parameters

Name	Type
«destructured»	ContextParams
onProgress?	(progress: number) => void

Returns

Promise<LlamaContext>

Defined in

index.ts:251

---

loadLlamaModelInfo

▸ loadLlamaModelInfo(model): Promise<Object>

Parameters

Name	Type
model	string

Returns

Promise<Object>

Defined in

index.ts:236

---

releaseAllLlama

▸ releaseAllLlama(): Promise<void>

Returns

Promise<void>

Defined in

index.ts:301

---

setContextLimit

▸ setContextLimit(limit): Promise<void>

Parameters

Name	Type
limit	number

Returns

Promise<void>

Defined in

index.ts:222