huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 10:38:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add support for the phi-hermes finetuned model. (#1192)

Browse Source
This commit is contained in:
Laurent Mazare
2023-10-27 05:57:08 +01:00
committed by GitHub
parent 0ec5ebcec4
commit 70d06ab4b0
2 changed files with 28 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
candle-examples/examples/phi/main.rs
View File

@ -124,6 +124,7 @@ enum WhichModel {
#[value(name = "1.5")] #[value(name = "1.5")]
V1_5, V1_5,
PuffinPhiV2, PuffinPhiV2,
PhiHermes,
} }
#[derive(Parser, Debug)] #[derive(Parser, Debug)]
@ -224,7 +225,9 @@ fn main() -> Result<()> {
match args.model { match args.model {
WhichModel::V1 => "microsoft/phi-1".to_string(), WhichModel::V1 => "microsoft/phi-1".to_string(),
WhichModel::V1_5 => "microsoft/phi-1_5".to_string(), WhichModel::V1_5 => "microsoft/phi-1_5".to_string(),
WhichModel::PuffinPhiV2 => "lmz/candle-quantized-phi".to_string(), WhichModel::PuffinPhiV2 | WhichModel::PhiHermes => {
"lmz/candle-quantized-phi".to_string()
}
} }
} }
} }
@ -238,7 +241,7 @@ fn main() -> Result<()> {
match args.model { match args.model {
WhichModel::V1 => "refs/pr/2".to_string(), WhichModel::V1 => "refs/pr/2".to_string(),
WhichModel::V1_5 => "refs/pr/18".to_string(), WhichModel::V1_5 => "refs/pr/18".to_string(),
WhichModel::PuffinPhiV2 => "main".to_string(), WhichModel::PuffinPhiV2 | WhichModel::PhiHermes => "main".to_string(),
} }
} }
} }
@ -248,7 +251,9 @@ fn main() -> Result<()> {
Some(file) => std::path::PathBuf::from(file), Some(file) => std::path::PathBuf::from(file),
None => match args.model { None => match args.model {
WhichModel::V1 | WhichModel::V1_5 => repo.get("tokenizer.json")?, WhichModel::V1 | WhichModel::V1_5 => repo.get("tokenizer.json")?,
WhichModel::PuffinPhiV2 => repo.get("tokenizer-puffin-phi-v2.json")?, WhichModel::PuffinPhiV2 | WhichModel::PhiHermes => {
repo.get("tokenizer-puffin-phi-v2.json")?
}
}, },
}; };
let filename = match args.weight_file { let filename = match args.weight_file {
@ -259,11 +264,13 @@ fn main() -> Result<()> {
WhichModel::V1 => repo.get("model-v1-q4k.gguf")?, WhichModel::V1 => repo.get("model-v1-q4k.gguf")?,
WhichModel::V1_5 => repo.get("model-q4k.gguf")?, WhichModel::V1_5 => repo.get("model-q4k.gguf")?,
WhichModel::PuffinPhiV2 => repo.get("model-puffin-phi-v2-q4k.gguf")?, WhichModel::PuffinPhiV2 => repo.get("model-puffin-phi-v2-q4k.gguf")?,
WhichModel::PhiHermes => repo.get("model-phi-hermes-1_3B-q4k.gguf")?,
} }
} else { } else {
match args.model { match args.model {
WhichModel::V1 | WhichModel::V1_5 => repo.get("model.safetensors")?, WhichModel::V1 | WhichModel::V1_5 => repo.get("model.safetensors")?,
WhichModel::PuffinPhiV2 => repo.get("model-puffin-phi-v2.safetensors")?, WhichModel::PuffinPhiV2 => repo.get("model-puffin-phi-v2.safetensors")?,
WhichModel::PhiHermes => repo.get("model-phi-hermes-1_3B.safetensors")?,
} }
} }
} }
@ -276,6 +283,7 @@ fn main() -> Result<()> {
WhichModel::V1 => Config::v1(), WhichModel::V1 => Config::v1(),
WhichModel::V1_5 => Config::v1_5(), WhichModel::V1_5 => Config::v1_5(),
WhichModel::PuffinPhiV2 => Config::puffin_phi_v2(), WhichModel::PuffinPhiV2 => Config::puffin_phi_v2(),
WhichModel::PhiHermes => Config::phi_hermes_1_3b(),
}; };
let (model, device) = if args.quantized { let (model, device) = if args.quantized {
let vb = candle_transformers::quantized_var_builder::VarBuilder::from_gguf(&filename)?; let vb = candle_transformers::quantized_var_builder::VarBuilder::from_gguf(&filename)?;

17
candle-transformers/src/models/mixformer.rs
View File

@ -73,6 +73,23 @@ impl Config {
pad_vocab_size_multiple: 64, pad_vocab_size_multiple: 64,
} }
} }
// https://huggingface.co/teknium/Phi-Hermes-1.3B/blob/main/config.json
pub fn phi_hermes_1_3b() -> Self {
Self {
vocab_size: 50304,
n_positions: 2048,
n_embd: 2048,
n_layer: 24,
n_inner: None,
n_head: 32,
rotary_dim: usize::min(32, 2048 / 32),
activation_function: Activation::NewGelu,
layer_norm_epsilon: 1e-5,
tie_word_embeddings: false,
pad_vocab_size_multiple: 64,
}
}
} }
#[derive(Debug, Clone)] #[derive(Debug, Clone)]