RGBCube/chatgpt.v
1
Fork 0
mirror of https://github.com/RGBCube/chatgpt.v synced 2025-07-27 16:57:46 +00:00
Code Issues Activity Actions

fix: value for the parameters of the config

Browse source
This commit is contained in:
RGBCube 2023-01-14 14:03:45 +03:00
parent a6181d5667
commit 645393d31c
2 changed files with 30 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/body.v
View file

@ -5,11 +5,11 @@ struct Body {
model string = 'text-davinci-003'
prompt string
max_tokens int
stop string
n u8
temperature f32
top_p f32
frequency_penalty f32
n u8
stop string
presence_penalty f32
frequency_penalty f32
best_of int
}

54
src/config.v
View file

@ -6,33 +6,25 @@ module chatgpt
pub struct GenerationConfig {
// This specifies the maximum number of tokens (common sequences of characters
// found in text) (basically words) ChatGPT should generate in its response.
max_tokens int = 256 // Min: 1, Max: 10240.
max_tokens u16 = 256 // Min: 1, Max: 10240.
// This specifies the level of "creativity" or "randomness" to use when
// generating text. A higher temperature will produce more varied and
// creative completions, while a lower temperature will produce more
// predictable and repetitive completions.
temperature f32 = 1 // Min: 0, Max: 2.
// This parameter is used to specify the fraction of the mass of the
// distribution to keep when selecting the next token. For example,
// if you set top_p to 0.4, ChatGPT will only consider the tokens
// with the highest probabilities (up to 40% of the total probability
// mass) when generating text. This can be used to produce more predictable
// and consistent completions (lower = more predictable), as the model will only select tokens from the
// most likely options.
top_p f32 = 1 // Min: 0, Max: 1.
// This specifies a sequence of characters that, when encountered by the
// model, will cause it to stop generating text. By default, it is set to
// none, which means that ChatGPT will not stop generating text until it
// reaches the maximum number of tokens specified by max_tokens.
stop ?string
// This specifies the level of "creativity" or "randomness" to use when
// generating text. A higher temperature will produce more varied and
// creative completions, while a lower temperature will produce more
// predictable and repetitive completions.
temperature f32 // Min: 0, Max: 2.
// This parameter is used to specify the fraction of the mass of the
// distribution to keep when selecting the next token. For example,
// if you set top_p to 0.5, ChatGPT will only consider the tokens
// with the highest probabilities (up to 50% of the total probability
// mass) when generating text. This can be used to produce more predictable
// and consistent completions, as the model will only select tokens from the
// most likely options.
top_p f32 = 1 // Min: 0, Max: 1.
// This parameter is used to specify a penalty to apply to the log probability
// of each token, based on how often it has been generated previously in
// the sequence. For example, if you set frequency_penalty to 0.1, ChatGPT
// will penalize tokens that have been generated more frequently in the
// sequence, making them less likely to be selected. This can be used to produce
// more diverse and interesting completions, as the model will avoid repeating
// the same tokens over and over.
frequency_penalty f32 // Min: 0, Max: 1.
// This parameter is used to specify a penalty to apply to the log probability
// of each token, based on how often it appears in the training data. For
// example, if you set presence_penalty to 0.1, the model will penalize
@ -40,18 +32,26 @@ pub struct GenerationConfig {
// to be selected. This can be used to produce more realistic and fluent
// completions, as the model will avoid generating rare or unusual tokens
// that do not appear often in real-world text.
presence_penalty f32 // Min: 0, Max: 1.
presence_penalty f32 // Min: -2.0, Max: 2.0.
// This parameter is used to specify a penalty to apply to the log probability
// of each token, based on how often it has been generated previously in
// the sequence. For example, if you set frequency_penalty to 0.1, ChatGPT
// will penalize tokens that have been generated more frequently in the
// sequence, making them less likely to be selected. This can be used to produce
// more diverse and interesting completions, as the model will avoid repeating
// the same tokens over and over.
frequency_penalty f32 // Min: -2.0, Max: 2.0.
// This parameter is used to specify the number of completions to generate
// for each prompt, and then return the highest-scoring completion(s). For example,
// if you set best_of to 3, the model will generate 3 completions for each prompt,
// and then return the highest-scoring completion(s). This can be useful if you want
// to ensure that the model returns the best possible completion(s) for each prompt.
best_of int = 1 // Min: 1, Max: 100.
best_of u8 = 1 // Min: 1, Max: 100.
}
// verify verifies that the SingularGenerationConfig is valid.
fn (c GenerationConfig) verify() ! {
if c.max_tokens < 1 || c.max_tokens > 10240 {
if c.max_tokens < 1 || c.max_tokens > 4096 {
return error('max_tokens must be between 1 and 10240')
}
if c.temperature < 0 || c.temperature > 2 {
@ -60,10 +60,10 @@ fn (c GenerationConfig) verify() ! {
if c.top_p < 0 || c.top_p > 1 {
return error('top_p must be between 0 and 1')
}
if c.frequency_penalty < 0 || c.frequency_penalty > 1 {
if c.frequency_penalty < -2 || c.frequency_penalty > 2 {
return error('frequency_penalty must be between 0 and 1')
}
if c.presence_penalty < 0 || c.presence_penalty > 1 {
if c.presence_penalty < -2 || c.presence_penalty > 2 {
return error('presence_penalty must be between 0 and 1')
}
if c.best_of < 1 || c.best_of > 100 {