Add the quantized mixformer model. (#953)

* Add the quantized mixformer model.

* Add the quantized option in the phi example.

candle-examples/examples/phi/main.rs

@@ -7,7 +7,8 @@ extern crate accelerate_src;
 use anyhow::{Error as E, Result};
 use clap::Parser;
 
-use candle_transformers::models::mixformer::{Config, MixFormerSequentialForCausalLM as Model};
+use candle_transformers::models::mixformer::{Config, MixFormerSequentialForCausalLM as MixFormer};
+use candle_transformers::models::quantized_mixformer::MixFormerSequentialForCausalLM as QMixFormer;
 
 use candle::{DType, Device, Tensor};
 use candle_nn::VarBuilder;
@@ -15,6 +16,11 @@ use candle_transformers::generation::LogitsProcessor;
 use hf_hub::{api::sync::Api, Repo, RepoType};
 use tokenizers::Tokenizer;
 
+enum Model {
+    MixFormer(MixFormer),
+    Quantized(QMixFormer),
+}
+
 struct TextGeneration {
     model: Model,
     device: Device,
@@ -58,7 +64,10 @@ impl TextGeneration {
             let context_size = if index > 0 { 1 } else { tokens.len() };
             let ctxt = &tokens[tokens.len().saturating_sub(context_size)..];
             let input = Tensor::new(ctxt, &self.device)?.unsqueeze(0)?;
-            let logits = self.model.forward(&input)?;
+            let logits = match &mut self.model {
+                Model::MixFormer(m) => m.forward(&input)?,
+                Model::Quantized(m) => m.forward(&input)?,
+            };
             let logits = logits.squeeze(0)?.to_dtype(DType::F32)?;
 
             let next_token = self.logits_processor.sample(&logits)?;
@@ -115,6 +124,9 @@ struct Args {
 
     #[arg(long)]
     weight_file: Option<String>,
+
+    #[arg(long)]
+    quantized: bool,
 }
 
 fn main() -> Result<()> {
@@ -150,10 +162,18 @@ fn main() -> Result<()> {
     let tokenizer = Tokenizer::from_file(tokenizer_filename).map_err(E::msg)?;
 
     let start = std::time::Instant::now();
+    let (model, device) = if args.quantized {
+        let vb = candle_transformers::quantized_var_builder::VarBuilder::from_gguf(&filenames[0])?;
+        let config = Config::v1_5();
+        let model = QMixFormer::new(&config, vb)?;
+        (Model::Quantized(model), Device::Cpu)
+    } else {
         let device = candle_examples::device(args.cpu)?;
         let vb = unsafe { VarBuilder::from_mmaped_safetensors(&filenames, DType::F32, &device)? };
         let config = Config::v1_5();
-    let model = Model::new(&config, vb)?;
+        let model = MixFormer::new(&config, vb)?;
+        (Model::MixFormer(model), device)
+    };
     println!("loaded the model in {:?}", start.elapsed());
 
     let mut pipeline = TextGeneration::new(

candle-transformers/src/models/mixformer.rs

@@ -10,17 +10,17 @@ const MAX_SEQ_LEN: usize = 4096;
 // https://huggingface.co/microsoft/phi-1_5/blob/main/configuration_mixformer_sequential.py
 #[derive(Debug, Clone, PartialEq)]
 pub struct Config {
-    vocab_size: usize,
+    pub(crate) vocab_size: usize,
-    n_positions: usize,
+    pub(crate) n_positions: usize,
-    n_embd: usize,
+    pub(crate) n_embd: usize,
-    n_layer: usize,
+    pub(crate) n_layer: usize,
-    n_inner: Option<usize>,
+    pub(crate) n_inner: Option<usize>,
-    n_head: usize,
+    pub(crate) n_head: usize,
-    rotary_dim: usize,
+    pub(crate) rotary_dim: usize,
-    activation_function: Activation,
+    pub(crate) activation_function: Activation,
-    layer_norm_epsilon: f64,
+    pub(crate) layer_norm_epsilon: f64,
-    tie_word_embeddings: bool,
+    pub(crate) tie_word_embeddings: bool,
-    pad_vocab_size_multiple: usize,
+    pub(crate) pad_vocab_size_multiple: usize,
 }
 
 impl Config {

candle-transformers/src/models/mod.rs

@@ -6,6 +6,7 @@ pub mod falcon;
 pub mod llama;
 pub mod mixformer;
 pub mod quantized_llama;
+pub mod quantized_mixformer;
 pub mod quantized_t5;
 pub mod segment_anything;
 pub mod stable_diffusion;

candle-transformers/src/models/quantized_mixformer.rs

@@ -0,0 +1,344 @@
+use crate::models::with_tracing::QMatMul;
+pub use crate::quantized_var_builder::VarBuilder;
+use candle::{DType, Device, IndexOp, Module, Result, Tensor, D};
+use candle_nn::Activation;
+
+pub use crate::models::mixformer::Config;
+
+const MAX_SEQ_LEN: usize = 4096;
+
+#[derive(Debug)]
+struct Embedding {
+    wte: super::quantized_t5::Embedding,
+}
+
+impl Embedding {
+    fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let wte = super::quantized_t5::Embedding::new(cfg.vocab_size, cfg.n_embd, vb.pp("wte"))?;
+        Ok(Self { wte })
+    }
+}
+
+impl Module for Embedding {
+    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
+        self.wte.forward(xs)
+    }
+}
+
+#[derive(Debug)]
+struct Linear {
+    weight: QMatMul,
+    bias: Option<Tensor>,
+}
+
+impl Module for Linear {
+    fn forward(&self, x: &Tensor) -> candle::Result<Tensor> {
+        let x = x.apply(&self.weight)?;
+        match &self.bias {
+            None => Ok(x),
+            Some(bias) => x.broadcast_add(bias),
+        }
+    }
+}
+
+fn linear(in_dim: usize, out_dim: usize, vb: VarBuilder) -> Result<Linear> {
+    let bias = vb.get(out_dim, "bias")?.dequantize(vb.device())?;
+    let weight = QMatMul::new(in_dim, out_dim, vb)?;
+    Ok(Linear {
+        weight,
+        bias: Some(bias),
+    })
+}
+
+fn layer_norm(size: usize, eps: f64, vb: VarBuilder) -> Result<candle_nn::LayerNorm> {
+    let weight = vb.get(size, "weight")?.dequantize(vb.device())?;
+    let bias = vb.get(size, "bias")?.dequantize(vb.device())?;
+    Ok(candle_nn::LayerNorm::new(weight, bias, eps))
+}
+
+fn get_mask(size: usize, device: &Device) -> Result<Tensor> {
+    let mask: Vec<_> = (0..size)
+        .flat_map(|i| (0..size).map(move |j| u8::from(j > i)))
+        .collect();
+    Tensor::from_slice(&mask, (size, size), device)
+}
+
+fn masked_fill(on_false: &Tensor, mask: &Tensor, on_true: f32) -> Result<Tensor> {
+    let shape = mask.shape();
+    let on_true = Tensor::new(on_true, on_false.device())?.broadcast_as(shape.dims())?;
+    let m = mask.where_cond(&on_true, on_false)?;
+    Ok(m)
+}
+
+#[derive(Debug)]
+struct RotaryEmbedding {
+    sin: Tensor,
+    cos: Tensor,
+}
+
+impl RotaryEmbedding {
+    fn new(dim: usize, max_seq_len: usize, dev: &Device) -> Result<Self> {
+        let inv_freq: Vec<_> = (0..dim)
+            .step_by(2)
+            .map(|i| 1f32 / 10000f32.powf(i as f32 / dim as f32))
+            .collect();
+        let inv_freq_len = inv_freq.len();
+        let inv_freq = Tensor::from_vec(inv_freq, (1, inv_freq_len), dev)?;
+        let t = Tensor::arange(0u32, max_seq_len as u32, dev)?
+            .to_dtype(DType::F32)?
+            .reshape((max_seq_len, 1))?;
+        let freqs = t.matmul(&inv_freq)?;
+        Ok(Self {
+            sin: freqs.sin()?,
+            cos: freqs.cos()?,
+        })
+    }
+
+    fn apply_rotary_emb_qkv(
+        &self,
+        qkv: &Tensor,
+        seqlen_offset: usize,
+    ) -> Result<(Tensor, Tensor, Tensor)> {
+        let (_b_size, seqlen, three, _, _headdim) = qkv.dims5()?;
+        if three != 3 {
+            candle::bail!("unexpected shape for qkv {:?}", qkv.shape())
+        }
+        let (_rotary_seqlen, rotary_dim) = self.cos.dims2()?;
+        let rotary_dim = rotary_dim * 2;
+        let q_rot = qkv.i((.., .., 0, .., ..rotary_dim))?;
+        let q_pass = qkv.i((.., .., 0, .., rotary_dim..))?;
+        let k_rot = qkv.i((.., .., 1, .., ..rotary_dim))?;
+        let k_pass = qkv.i((.., .., 1, .., rotary_dim..))?;
+        let q12 = q_rot.chunk(2, D::Minus1)?;
+        let k12 = k_rot.chunk(2, D::Minus1)?;
+        let (q1, q2) = (&q12[0], &q12[1]);
+        let (k1, k2) = (&k12[0], &k12[1]);
+        let c = self.cos.narrow(0, seqlen_offset, seqlen)?.unsqueeze(1)?;
+        let s = self.sin.narrow(0, seqlen_offset, seqlen)?.unsqueeze(1)?;
+        let q_rot = Tensor::cat(
+            &[
+                (q1.broadcast_mul(&c)? - q2.broadcast_mul(&s)?)?,
+                (q1.broadcast_mul(&s)? + q2.broadcast_mul(&c)?)?,
+            ],
+            D::Minus1,
+        )?;
+        let k_rot = Tensor::cat(
+            &[
+                (k1.broadcast_mul(&c)? - k2.broadcast_mul(&s)?)?,
+                (k1.broadcast_mul(&s)? + k2.broadcast_mul(&c)?)?,
+            ],
+            D::Minus1,
+        )?;
+        let q = Tensor::cat(&[&q_rot, &q_pass], D::Minus1)?;
+        let k = Tensor::cat(&[&k_rot, &k_pass], D::Minus1)?;
+        let v = qkv.i((.., .., 2))?;
+        Ok((q, k, v))
+    }
+}
+
+#[derive(Debug)]
+#[allow(clippy::upper_case_acronyms)]
+struct MLP {
+    fc1: Linear,
+    fc2: Linear,
+    act: Activation,
+}
+
+impl MLP {
+    fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let n_inner = cfg.n_inner.unwrap_or(4 * cfg.n_embd);
+        let fc1 = linear(cfg.n_embd, n_inner, vb.pp("fc1"))?;
+        let fc2 = linear(n_inner, cfg.n_embd, vb.pp("fc2"))?;
+        Ok(Self {
+            fc1,
+            fc2,
+            act: cfg.activation_function,
+        })
+    }
+}
+
+impl Module for MLP {
+    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
+        xs.apply(&self.fc1)?.apply(&self.act)?.apply(&self.fc2)
+    }
+}
+
+#[derive(Debug)]
+struct CausalLMHead {
+    ln: candle_nn::LayerNorm,
+    linear: Linear,
+}
+
+impl CausalLMHead {
+    fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let ln = layer_norm(cfg.n_embd, cfg.layer_norm_epsilon, vb.pp("ln"))?;
+        let linear = linear(cfg.n_embd, cfg.vocab_size, vb.pp("linear"))?;
+        Ok(Self { ln, linear })
+    }
+}
+
+impl Module for CausalLMHead {
+    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
+        xs.apply(&self.ln)?
+            .apply(&self.linear)?
+            .to_dtype(DType::F32)
+    }
+}
+
+#[derive(Debug)]
+#[allow(clippy::upper_case_acronyms)]
+struct MHA {
+    wqkv: Linear,
+    out_proj: Linear,
+    rotary_emb: RotaryEmbedding,
+    kv_cache: Option<(Tensor, Tensor)>,
+    head_dim: usize,
+    n_head: usize,
+    softmax_scale: f64,
+    span: tracing::Span,
+}
+
+impl MHA {
+    fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let head_dim = cfg.n_embd / cfg.n_head;
+        let op_size = cfg.n_embd;
+        let wqkv = linear(cfg.n_embd, 3 * op_size, vb.pp("Wqkv"))?;
+        let out_proj = linear(op_size, cfg.n_embd, vb.pp("out_proj"))?;
+        let rotary_emb = RotaryEmbedding::new(cfg.rotary_dim, MAX_SEQ_LEN, vb.device())?;
+        let softmax_scale = 1f64 / (head_dim as f64).sqrt();
+        Ok(Self {
+            wqkv,
+            out_proj,
+            head_dim,
+            n_head: cfg.n_head,
+            kv_cache: None,
+            rotary_emb,
+            softmax_scale,
+            span: tracing::span!(tracing::Level::TRACE, "mha"),
+        })
+    }
+
+    fn forward(&mut self, xs: &Tensor, mask: Option<&Tensor>) -> Result<Tensor> {
+        let _enter = self.span.enter();
+        let (b_size, seq_len, _n_embd) = xs.dims3()?;
+        let qkv = self
+            .wqkv
+            .forward(xs)?
+            .reshape((b_size, seq_len, 3, (), self.head_dim))?;
+        let seqlen_offset = match &self.kv_cache {
+            None => 0,
+            Some((prev_k, _)) => prev_k.dim(1)?,
+        };
+        // In the python implementation, a single tensor is returned with the third axis of size 3.
+        let (q, k, v) = self.rotary_emb.apply_rotary_emb_qkv(&qkv, seqlen_offset)?;
+        let (k, v) = match &self.kv_cache {
+            None => (k, v),
+            Some((prev_k, prev_v)) => {
+                let k = Tensor::cat(&[prev_k, &k], 1)?;
+                let v = Tensor::cat(&[prev_v, &v], 1)?;
+                (k, v)
+            }
+        };
+        self.kv_cache = Some((k.clone(), v.clone()));
+        // scores = torch.einsum('bthd,bshd->bhts', q, k * softmax_scale)
+        let q = q.transpose(1, 2)?.flatten_to(1)?; // b*h, t, d
+        let k = k.transpose(1, 2)?.flatten_to(1)?; // b*h, s, d
+        let v = v.transpose(1, 2)?.flatten_to(1)?; // b*h, s, d
+        let attn_weights = (q.matmul(&k.t()?)? * self.softmax_scale)?; // b*h, t, s
+
+        // causal_mask = torch.triu(torch.full((seqlen_q, seqlen_k), -10000.0, device=scores.device), 1)
+        // scores = scores + causal_mask.to(dtype=scores.dtype)
+        let attn_weights = match mask {
+            None => attn_weights,
+            Some(mask) => masked_fill(
+                &attn_weights,
+                &mask.broadcast_left(b_size * self.n_head)?,
+                f32::NEG_INFINITY,
+            )?,
+        };
+        let attn_weights = candle_nn::ops::softmax(&attn_weights, D::Minus1)?;
+
+        // output = torch.einsum('bhts,bshd->bthd', attention_drop, v)
+        // attn_weights: b*h,t,s, v: b*h,s,d
+        let attn_output = attn_weights.matmul(&v)?;
+        // b*h,t,d
+        let attn_output = attn_output
+            .reshape((b_size, (), seq_len, self.head_dim))?
+            .transpose(1, 2)?
+            .flatten_from(D::Minus2)?;
+        attn_output.apply(&self.out_proj)
+    }
+}
+
+#[derive(Debug)]
+struct ParallelBlock {
+    ln: candle_nn::LayerNorm,
+    mixer: MHA,
+    mlp: MLP,
+    span: tracing::Span,
+}
+
+impl ParallelBlock {
+    fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let ln = layer_norm(cfg.n_embd, cfg.layer_norm_epsilon, vb.pp("ln"))?;
+        let mixer = MHA::new(cfg, vb.pp("mixer"))?;
+        let mlp = MLP::new(cfg, vb.pp("mlp"))?;
+        Ok(Self {
+            ln,
+            mixer,
+            mlp,
+            span: tracing::span!(tracing::Level::TRACE, "block"),
+        })
+    }
+
+    fn forward(&mut self, xs: &Tensor, mask: Option<&Tensor>) -> Result<Tensor> {
+        let _enter = self.span.enter();
+        let residual = xs;
+        let xs = xs.apply(&self.ln)?;
+        let attn_outputs = self.mixer.forward(&xs, mask)?;
+        let feed_forward_hidden_states = self.mlp.forward(&xs)?;
+        attn_outputs + feed_forward_hidden_states + residual
+    }
+}
+
+#[derive(Debug)]
+pub struct MixFormerSequentialForCausalLM {
+    embedding: Embedding,
+    blocks: Vec<ParallelBlock>,
+    head: CausalLMHead,
+    span: tracing::Span,
+}
+
+impl MixFormerSequentialForCausalLM {
+    pub fn new(cfg: &Config, vb: VarBuilder) -> Result<Self> {
+        let vb = vb.pp("layers");
+        let embedding = Embedding::new(cfg, vb.pp(0))?;
+        let mut blocks = Vec::new();
+        for i in 0..cfg.n_layer {
+            let block = ParallelBlock::new(cfg, vb.pp(i + 1))?;
+            blocks.push(block)
+        }
+        let head = CausalLMHead::new(cfg, vb.pp(cfg.n_layer + 1))?;
+        Ok(Self {
+            embedding,
+            blocks,
+            head,
+            span: tracing::span!(tracing::Level::TRACE, "mixformer"),
+        })
+    }
+
+    pub fn forward(&mut self, xs: &Tensor) -> Result<Tensor> {
+        let _enter = self.span.enter();
+        let (_b_size, seq_len) = xs.dims2()?;
+        let mut xs = xs.apply(&self.embedding)?;
+        let mask = if seq_len <= 1 {
+            None
+        } else {
+            Some(get_mask(seq_len, xs.device())?)
+        };
+        for block in self.blocks.iter_mut() {
+            xs = block.forward(&xs, mask.as_ref())?
+        }
+        xs.narrow(1, seq_len - 1, 1)?.apply(&self.head)?.squeeze(1)
+    }
+}

candle-transformers/src/models/quantized_t5.rs

@@ -1,6 +1,7 @@
 // T5 Text Model, quantized version
 // https://github.com/huggingface/transformers/blob/main/src/transformers/models/t5/modeling_t5.py
 
+use crate::models::with_tracing::QMatMul;
 pub use crate::quantized_var_builder::VarBuilder;
 use candle::{DType, Device, Module, Result, Tensor, D};
 use candle_nn::Activation;
@@ -8,20 +9,20 @@ use serde::Deserialize;
 use std::sync::Arc;
 
 #[derive(Debug)]
-struct Embedding {
+pub struct Embedding {
     inner: candle_nn::Embedding,
     span: tracing::Span,
 }
 
 impl Embedding {
-    fn new(d1: usize, d2: usize, vb: VarBuilder) -> Result<Self> {
+    pub fn new(d1: usize, d2: usize, vb: VarBuilder) -> Result<Self> {
         let embeddings = vb.get((d1, d2), "weight")?.dequantize(vb.device())?;
         let inner = candle_nn::Embedding::new(embeddings, d2);
         let span = tracing::span!(tracing::Level::TRACE, "embedding");
         Ok(Self { inner, span })
     }
 
-    fn embeddings(&self) -> &Tensor {
+    pub fn embeddings(&self) -> &Tensor {
         self.inner.embeddings()
     }
 }
@@ -33,34 +34,6 @@ impl Module for Embedding {
     }
 }
 
-// QMatMul wrapper adding some tracing.
-struct QMatMul {
-    inner: candle::quantized::QMatMul,
-    span: tracing::Span,
-}
-
-impl QMatMul {
-    fn new(out_dim: usize, in_dim: usize, vb: VarBuilder) -> Result<Self> {
-        let ws = vb.get((in_dim, out_dim), "weight")?;
-        let inner = candle::quantized::QMatMul::from_arc(ws);
-        let span = tracing::span!(tracing::Level::TRACE, "qmatmul");
-        Ok(Self { inner, span })
-    }
-}
-
-impl Module for QMatMul {
-    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
-        let _enter = self.span.enter();
-        self.inner.forward(xs)
-    }
-}
-
-impl std::fmt::Debug for QMatMul {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "QMatMul")
-    }
-}
-
 fn default_relative_attention_max_distance() -> usize {
     128
 }

candle-transformers/src/models/with_tracing.rs

@@ -76,3 +76,35 @@ pub fn conv2d(
     let inner = candle_nn::conv2d(in_channels, out_channels, kernel_size, cfg, vs)?;
     Ok(Conv2d { inner, span })
 }
+
+// QMatMul wrapper adding some tracing.
+pub struct QMatMul {
+    inner: candle::quantized::QMatMul,
+    span: tracing::Span,
+}
+
+impl QMatMul {
+    pub fn new(
+        out_dim: usize,
+        in_dim: usize,
+        vb: crate::quantized_var_builder::VarBuilder,
+    ) -> Result<Self> {
+        let ws = vb.get((in_dim, out_dim), "weight")?;
+        let inner = candle::quantized::QMatMul::from_arc(ws);
+        let span = tracing::span!(tracing::Level::TRACE, "qmatmul");
+        Ok(Self { inner, span })
+    }
+}
+
+impl Module for QMatMul {
+    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
+        let _enter = self.span.enter();
+        self.inner.forward(xs)
+    }
+}
+
+impl std::fmt::Debug for QMatMul {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "QMatMul")
+    }
+}