Merge pull request #276 from EricLBuehler/quantized_phi3

Add the quantized phi3 model

mistralrs-core/src/layers.rs

@@ -27,6 +27,15 @@ impl RmsNorm {
             weight: w,
         })
     }
+
+    pub fn from_w(w: Tensor, eps: f64) -> Result<Self> {
+        let inner = candle_nn::RmsNorm::<RmsNormNonQuantized>::new(w.clone(), eps);
+        Ok(Self {
+            inner,
+            eps,
+            weight: w,
+        })
+    }
 }
 
 impl Module for RmsNorm {

mistralrs-core/src/models/mod.rs

@@ -12,6 +12,7 @@ pub(crate) mod phi2;
 pub(crate) mod phi3;
 pub(crate) mod quantized_llama;
 pub(crate) mod quantized_phi2;
+pub(crate) mod quantized_phi3;
 pub(crate) mod qwen2;
 
 pub type LayerCaches = Vec<Option<(Tensor, Tensor)>>;

mistralrs-core/src/models/quantized_phi3.rs

@@ -0,0 +1,331 @@
+#![allow(clippy::cast_possible_truncation, clippy::cast_precision_loss)]
+
+use std::collections::HashMap;
+
+use crate::device_map::DeviceMapper;
+use crate::layers::RmsNorm;
+use crate::DeviceMapMetadata;
+use candle_core::quantized::gguf_file;
+use candle_core::quantized::QMatMul;
+use candle_core::quantized::QTensor;
+use candle_core::{DType, Device, IndexOp, Module, Result, Tensor, D};
+use candle_nn::Embedding;
+
+use super::repeat_kv;
+use super::verify_sanity_gguf;
+use super::Cache;
+
+#[derive(Debug, Clone)]
+struct Mlp {
+    ffn_up: QMatMul,
+    ffn_down: QMatMul,
+    i_size: usize,
+}
+
+impl Module for Mlp {
+    fn forward(&self, xs: &Tensor) -> Result<Tensor> {
+        let up_states = xs.apply(&self.ffn_up)?;
+        let gate = up_states.narrow(D::Minus1, 0, self.i_size)?;
+        let up_states = up_states.narrow(D::Minus1, self.i_size, self.i_size)?;
+        let up_states = (up_states * gate.silu()?)?;
+        up_states.apply(&self.ffn_down)
+    }
+}
+
+fn rms_norm(w: QTensor, eps: f64) -> Result<RmsNorm> {
+    let w = w.dequantize(&w.device())?;
+    let rms = RmsNorm::from_w(w, eps)?;
+    Ok(rms)
+}
+
+#[derive(Debug, Clone)]
+struct LayerWeights {
+    attn_qkv: QMatMul,
+    attn_output: QMatMul,
+    attn_norm: RmsNorm,
+    ffn_norm: RmsNorm,
+    mlp: Mlp,
+    n_head: usize,
+    n_kv_head: usize,
+    head_dim: usize,
+    cos: Tensor,
+    sin: Tensor,
+    neg_inf: Tensor,
+    sliding_window: usize,
+}
+
+fn masked_fill(on_false: &Tensor, mask: &Tensor, on_true: &Tensor) -> Result<Tensor> {
+    let shape = mask.shape();
+    let m = mask.where_cond(&on_true.broadcast_as(shape.dims())?, on_false)?;
+    Ok(m)
+}
+
+impl LayerWeights {
+    fn apply_rotary_emb(&self, xs: &Tensor, seqlen_offsets: &[usize]) -> Result<Tensor> {
+        let (_b_sz, _h, seq_len, _n_embd) = xs.dims4()?;
+        let mut outputs = Vec::new();
+        for (i, offset) in seqlen_offsets.iter().enumerate() {
+            let cos = self.cos.narrow(0, *offset, seq_len)?;
+            let sin = self.sin.narrow(0, *offset, seq_len)?;
+            outputs.push(candle_nn::rotary_emb::rope(
+                &xs.i(i)?.unsqueeze(0)?.contiguous()?,
+                &cos,
+                &sin,
+            )?);
+        }
+        Tensor::cat(&outputs, 0)
+    }
+
+    fn forward_attn(
+        &mut self,
+        x: &Tensor,
+        mask: Option<&Tensor>,
+        seqlen_offsets: &[usize],
+        kv_cache: &mut Option<(Tensor, Tensor)>,
+    ) -> Result<Tensor> {
+        let (b_sz, seq_len, n_embd) = x.dims3()?;
+        let qkv = self.attn_qkv.forward(x)?;
+
+        let query_pos = self.n_head * self.head_dim;
+        let q = qkv.narrow(D::Minus1, 0, query_pos)?;
+        let k = qkv.narrow(D::Minus1, query_pos, self.n_kv_head * self.head_dim)?;
+        let v = qkv.narrow(
+            D::Minus1,
+            query_pos + self.n_kv_head * self.head_dim,
+            self.n_kv_head * self.head_dim,
+        )?;
+
+        let q = q
+            .reshape((b_sz, seq_len, self.n_head, self.head_dim))?
+            .transpose(1, 2)?;
+        let k = k
+            .reshape((b_sz, seq_len, self.n_head, self.head_dim))?
+            .transpose(1, 2)?;
+        let v = v
+            .reshape((b_sz, seq_len, self.n_kv_head, self.head_dim))?
+            .transpose(1, 2)?;
+
+        let q = self.apply_rotary_emb(&q, seqlen_offsets)?.contiguous()?;
+        let k = self.apply_rotary_emb(&k, seqlen_offsets)?;
+
+        let (k, v, attn_mask) = match kv_cache.clone() {
+            None => (k, v, mask.cloned()),
+            Some((mut prev_k, mut prev_v)) => {
+                let mut mask = mask.cloned();
+                let kv_seq_len = prev_k.dim(2)?;
+                let sliding_window = self.sliding_window;
+                if kv_seq_len > sliding_window {
+                    prev_k =
+                        prev_k.narrow(2, kv_seq_len - (sliding_window - 1), sliding_window - 1)?;
+                    prev_v =
+                        prev_v.narrow(2, kv_seq_len - (sliding_window - 1), sliding_window - 1)?;
+                    if let Some(ref mut mask) = mask {
+                        let mask_len = mask.dim(1)?;
+                        *mask =
+                            mask.narrow(1, mask_len - (sliding_window - 1), sliding_window - 1)?;
+                        *mask = Tensor::cat(
+                            &[&*mask, &mask.narrow(1, mask_len - 1, 1)?.ones_like()?],
+                            D::Minus1,
+                        )?;
+                    }
+                }
+                let k = Tensor::cat(&[prev_k, k], 2)?;
+                let v = Tensor::cat(&[prev_v, v], 2)?;
+                (k, v, mask)
+            }
+        };
+        *kv_cache = Some((k.clone(), v.clone()));
+
+        let k = repeat_kv(k, self.n_head / self.n_kv_head)?;
+        let v = repeat_kv(v, self.n_head / self.n_kv_head)?;
+
+        let att = (q.matmul(&k.t()?)? / (self.head_dim as f64).sqrt())?;
+        let att = match attn_mask {
+            None => att,
+            Some(mask) => {
+                let mask = mask.broadcast_as(att.shape())?;
+                masked_fill(&att, &mask, &self.neg_inf)?
+            }
+        };
+        let att = candle_nn::ops::softmax_last_dim(&att)?;
+        // Convert to contiguous as matmul doesn't support strided vs for now.
+        let y = att.matmul(&v.contiguous()?)?;
+        let y = y.transpose(1, 2)?.reshape(&[b_sz, seq_len, n_embd])?;
+        let y = self.attn_output.forward(&y)?;
+        Ok(y)
+    }
+}
+
+#[derive(Debug)]
+pub struct ModelWeights {
+    tok_embeddings: Embedding,
+    layers: Vec<LayerWeights>,
+    output_norm: RmsNorm,
+    output: QMatMul,
+    masks: HashMap<usize, Tensor>,
+    mapper: Option<Box<dyn DeviceMapper + Send + Sync>>,
+    pub device: Device,
+    pub cache: Cache,
+    pub max_seq_len: usize,
+}
+
+fn precomput_freqs_cis(
+    head_dim: usize,
+    freq_base: f32,
+    device: &Device,
+    context_window: usize,
+) -> Result<(Tensor, Tensor)> {
+    let theta: Vec<_> = (0..head_dim)
+        .step_by(2)
+        .map(|i| 1f32 / freq_base.powf(i as f32 / head_dim as f32))
+        .collect();
+    let theta = Tensor::new(theta.as_slice(), device)?;
+    let idx_theta = Tensor::arange(0, context_window as u32, device)?
+        .to_dtype(DType::F32)?
+        .reshape((context_window, 1))?
+        .matmul(&theta.reshape((1, theta.elem_count()))?)?;
+    let cos = idx_theta.cos()?;
+    let sin = idx_theta.sin()?;
+    Ok((cos, sin))
+}
+
+impl ModelWeights {
+    pub fn from_gguf<R: std::io::Seek + std::io::Read>(
+        ct: gguf_file::Content,
+        reader: &mut R,
+        device: &Device,
+        mapper: DeviceMapMetadata,
+    ) -> Result<Self> {
+        let md_get = |s: &str| match ct.metadata.get(s) {
+            None => candle_core::bail!("cannot find {s} in metadata"),
+            Some(v) => Ok(v),
+        };
+        verify_sanity_gguf(md_get("general.architecture")?.to_string().unwrap(), "phi3")?;
+
+        // Parameter extraction from metadata.
+        let head_count = md_get("phi3.attention.head_count")?.to_u32()? as usize;
+        let head_count_kv = md_get("phi3.attention.head_count_kv")?.to_u32()? as usize;
+        let block_count = md_get("phi3.block_count")?.to_u32()? as usize;
+        let embedding_length = md_get("phi3.embedding_length")?.to_u32()? as usize;
+        let i_size = md_get("phi3.feed_forward_length")?.to_u32()? as usize;
+        let rope_dim = md_get("phi3.rope.dimension_count")?.to_u32()? as usize;
+        let rms_eps = md_get("phi3.attention.layer_norm_rms_epsilon")?.to_f32()? as f64;
+        let context_window = md_get("phi3.context_length")?.to_u32()? as usize;
+        let (cos, sin) = precomput_freqs_cis(rope_dim, 10_000., device, context_window)?;
+        let neg_inf = Tensor::new(f32::NEG_INFINITY, device)?;
+
+        let tok_embeddings = ct.tensor(reader, "token_embd.weight", device)?;
+        let tok_embeddings = tok_embeddings.dequantize(device)?;
+        let output_norm = rms_norm(ct.tensor(reader, "output_norm.weight", device)?, rms_eps)?;
+        let output = QMatMul::from_qtensor(ct.tensor(reader, "output.weight", device)?)?;
+        let mut layers = Vec::with_capacity(block_count);
+        let mapper = mapper.into_mapper(block_count, device)?;
+        for layer_idx in 0..block_count {
+            let prefix = format!("blk.{layer_idx}");
+            let device = mapper.device_for(layer_idx, false).unwrap_or(device);
+            let ffn_up = QMatMul::from_qtensor(ct.tensor(
+                reader,
+                &format!("{prefix}.ffn_up.weight"),
+                device,
+            )?)?;
+            let ffn_down = QMatMul::from_qtensor(ct.tensor(
+                reader,
+                &format!("{prefix}.ffn_down.weight"),
+                device,
+            )?)?;
+            let mlp = Mlp {
+                ffn_up,
+                ffn_down,
+                i_size,
+            };
+            let attn_norm = rms_norm(
+                ct.tensor(reader, &format!("{prefix}.attn_norm.weight"), device)?,
+                rms_eps,
+            )?;
+            let ffn_norm = rms_norm(
+                ct.tensor(reader, &format!("{prefix}.ffn_norm.weight"), device)?,
+                rms_eps,
+            )?;
+            layers.push(LayerWeights {
+                attn_qkv: QMatMul::from_qtensor(ct.tensor(
+                    reader,
+                    &format!("{prefix}.attn_qkv.weight"),
+                    device,
+                )?)?,
+                attn_output: QMatMul::from_qtensor(ct.tensor(
+                    reader,
+                    &format!("{prefix}.attn_output.weight"),
+                    device,
+                )?)?,
+                attn_norm,
+                ffn_norm,
+                mlp,
+                n_head: head_count,
+                n_kv_head: head_count_kv,
+                head_dim: embedding_length / head_count,
+                cos: cos.clone(),
+                sin: sin.clone(),
+                neg_inf: neg_inf.clone(),
+                sliding_window: context_window,
+            })
+        }
+        Ok(Self {
+            tok_embeddings: Embedding::new(tok_embeddings, embedding_length),
+            layers,
+            output_norm,
+            output,
+            masks: HashMap::new(),
+            mapper: Some(mapper),
+            device: device.clone(),
+            cache: Cache::new(block_count, false),
+            max_seq_len: context_window,
+        })
+    }
+
+    fn mask(&mut self, t: usize, device: &Device) -> Result<Tensor> {
+        if let Some(mask) = self.masks.get(&t) {
+            Ok(mask.clone())
+        } else {
+            let mask: Vec<_> = (0..t)
+                .flat_map(|i| (0..t).map(move |j| u8::from(j > i)))
+                .collect();
+            let mask = Tensor::from_slice(&mask, (t, t), device)?;
+            self.masks.insert(t, mask.clone());
+            Ok(mask)
+        }
+    }
+
+    pub fn forward(&mut self, xs: &Tensor, seqlen_offsets: &[usize]) -> Result<Tensor> {
+        let (_b_sz, seq_len) = xs.dims2()?;
+        let mask = if seq_len == 1 {
+            None
+        } else {
+            Some(self.mask(seq_len, xs.device())?)
+        };
+        let mut xs = self.tok_embeddings.forward(xs)?;
+        let mut cache = self.cache.lock();
+        for (i, layer) in self.layers.iter_mut().enumerate() {
+            if let Some(ref mapper) = self.mapper {
+                xs = mapper.map(xs, i)?;
+            }
+            let residual = &xs;
+            let ys = xs.apply(&layer.attn_norm)?;
+            let ys = layer.forward_attn(
+                &ys,
+                mask.as_ref()
+                    .map(|m| m.to_device(xs.device()).unwrap())
+                    .as_ref(),
+                seqlen_offsets,
+                &mut cache[i],
+            )?;
+            let ys = (ys + residual)?;
+            let residual = &ys;
+            let ys = ys.apply(&layer.ffn_norm)?;
+            let ys = layer.mlp.forward(&ys)?;
+            xs = (ys + residual)?
+        }
+        let xs = xs.to_device(&self.device)?;
+        let xs = xs.apply(&self.output_norm)?.i((.., seq_len - 1, ..))?;
+        self.output.forward(&xs)
+    }
+}