feat: mixtral (#1328)

Dockerfile

@@ -154,6 +154,11 @@ COPY server/Makefile-vllm Makefile
 # Build specific version of vllm
 RUN make build-vllm-cuda
 
+# Build megablocks
+FROM kernel-builder as megablocks-builder
+
+RUN pip install git+https://github.com/OlivierDehaene/megablocks@181709df192de9a941fdf3a641cdc65a0462996e
+
 # Text Generation Inference base image
 FROM nvidia/cuda:12.1.0-base-ubuntu20.04 as base
 
@@ -175,8 +180,8 @@ RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-ins
         curl \
         && rm -rf /var/lib/apt/lists/*
 
-# Copy conda with PyTorch installed
-COPY --from=pytorch-install /opt/conda /opt/conda
+# Copy conda with PyTorch and Megablocks installed
+COPY --from=megablocks-builder /opt/conda /opt/conda
 
 # Copy build artifacts from flash attention builder
 COPY --from=flash-att-builder /usr/src/flash-attention/build/lib.linux-x86_64-cpython-310 /opt/conda/lib/python3.10/site-packages

router/src/server.rs

@@ -629,6 +629,9 @@ pub async fn run(
     // Batch size buckets
     let batch_size_matcher = Matcher::Full(String::from("tgi_batch_next_size"));
     let batch_size_buckets: Vec<f64> = (0..1024).map(|x| (x + 1) as f64).collect();
+    // Speculated tokens buckets
+    let skipped_matcher = Matcher::Full(String::from("tgi_request_skipped_tokens"));
+    let skipped_buckets: Vec<f64> = (0..shard_info.speculate + 1).map(|x| x as f64).collect();
 
     // Prometheus handler
     let builder = PrometheusBuilder::new()
@@ -641,6 +644,8 @@ pub async fn run(
         .set_buckets_for_metric(max_new_tokens_matcher, &max_new_tokens_buckets)
         .unwrap()
         .set_buckets_for_metric(batch_size_matcher, &batch_size_buckets)
+        .unwrap()
+        .set_buckets_for_metric(skipped_matcher, &skipped_buckets)
         .unwrap();
     let prom_handle = builder
         .install_recorder()

server/Makefile

@@ -16,6 +16,9 @@ gen-server:
 	find text_generation_server/pb/ -type f -name "*.py" -print0 -exec sed -i -e 's/^\(import.*pb2\)/from . \1/g' {} \;
 	touch text_generation_server/pb/__init__.py
 
+install-megablocks:
+	pip install git+https://github.com/OlivierDehaene/megablocks@181709df192de9a941fdf3a641cdc65a0462996e
+
 install: gen-server
 	pip install pip --upgrade
 	pip install -r requirements_cuda.txt

server/text_generation_server/models/__init__.py

@@ -1,4 +1,3 @@
-import os
 import torch
 
 from loguru import logger
@@ -78,6 +77,18 @@ except ImportError as e:
 if MISTRAL:
     __all__.append(FlashMistral)
 
+MIXTRAL = True
+try:
+    from text_generation_server.models.flash_mixtral import FlashMixtral
+except ImportError as e:
+    logger.warning(f"Could not import Mixtral model: {e}")
+    MIXTRAL = False
+
+if MIXTRAL:
+    __all__.append(FlashMixtral)
+
+
+
 def get_model(
     model_id: str,
     revision: Optional[str],
@@ -141,7 +152,6 @@ def get_model(
     use_medusa = None
     if "medusa_num_heads" in config_dict:
         use_medusa = model_id
-        medusa_config = config_dict
         model_id = config_dict["base_model_name_or_path"]
         revision = "main"
         speculate_medusa = config_dict["medusa_num_heads"]
@@ -292,7 +302,18 @@ def get_model(
                 dtype=dtype,
                 trust_remote_code=trust_remote_code,
             )
-        raise NotImplementedError("Mistral model requires flash attention v2")
+        raise NotImplementedError("Mistral models requires flash attention v2")
+
+    if model_type == "mixtral":
+        if MIXTRAL:
+            return FlashMixtral(
+                model_id,
+                revision,
+                quantize=quantize,
+                dtype=dtype,
+                trust_remote_code=trust_remote_code,
+            )
+        raise NotImplementedError("Mixtral models requires flash attention v2, stk and megablocks")
 
     if model_type == "opt":
         return OPTSharded(

server/text_generation_server/models/custom_modeling/flash_llama_modeling.py

@@ -34,14 +34,8 @@ from text_generation_server.utils.layers import (
     PositionRotaryEmbedding,
     TensorParallelHead,
     get_linear,
+    FastRMSNorm
 )
-from text_generation_server.utils.import_utils import IS_CUDA_SYSTEM, IS_ROCM_SYSTEM
-
-if IS_CUDA_SYSTEM:
-    import dropout_layer_norm
-elif IS_ROCM_SYSTEM:
-    from vllm import layernorm_ops
-
 
 class LlamaConfig(PretrainedConfig):
     def __init__(
@@ -95,75 +89,6 @@ class LlamaConfig(PretrainedConfig):
         )
 
 
-class LlamaRMSNorm(nn.Module):
-    def __init__(self, prefix, weights, eps=1e-6):
-        """
-        LlamaRMSNorm is equivalent to T5LayerNorm
-        """
-        super().__init__()
-
-        weight = weights.get_tensor(f"{prefix}.weight")
-        self.weight = nn.Parameter(weight)
-        self.variance_epsilon = eps
-
-    def forward(self, hidden_states, residual=None):
-        if hidden_states.shape[-1] > 8192:
-            if residual is not None:
-                hidden_states += residual
-            residual = hidden_states
-
-            hidden_states = hidden_states.to(torch.float32)
-            variance = hidden_states.pow(2).mean(-1, keepdim=True)
-            hidden_states = hidden_states * torch.rsqrt(
-                variance + self.variance_epsilon
-            )
-
-            # convert into half-precision if necessary
-            if self.weight.dtype in [torch.float16, torch.bfloat16]:
-                hidden_states = hidden_states.to(self.weight.dtype)
-
-            return self.weight * hidden_states, residual
-        elif IS_CUDA_SYSTEM:
-            # faster post attention rms norm
-            normed_hidden_states, res, *rest = dropout_layer_norm.dropout_add_ln_fwd(
-                hidden_states,
-                residual,
-                self.weight,
-                None,
-                None,
-                None,
-                None,
-                None,
-                0.0,
-                self.variance_epsilon,
-                1.0,
-                0,
-                None,
-                False,
-                True,  # Activate RMSNorm
-            )
-            if res is None:
-                res = hidden_states
-
-            return normed_hidden_states, res
-        elif IS_ROCM_SYSTEM:
-            # We use VLLM RMSNorm kernel that can be compiled for RoCm, instead of Flash Attention ones that can not.
-            if residual is not None:
-                hidden_states += residual
-            residual = hidden_states
-
-            out = torch.empty_like(hidden_states)
-            layernorm_ops.rms_norm(
-                out,
-                hidden_states,
-                self.weight.data,
-                self.variance_epsilon,
-            )
-            return out, residual
-        else:
-            raise ValueError("Your system seem to be not supported. Please check your install or open an issue at https://github.com/huggingface/text-generation-inference/issues with a clear reproduction.")
-
-
 def load_attention(config, prefix, weights):
     if config.num_attention_heads != config.num_key_value_heads:
         return _load_gqa(config, prefix, weights)
@@ -363,10 +288,8 @@ class FlashLlamaLayer(nn.Module):
         )
         self.mlp = LlamaMLP(prefix=f"{prefix}.mlp", config=config, weights=weights)
 
-        self.input_layernorm = LlamaRMSNorm(
-            prefix=f"{prefix}.input_layernorm", weights=weights, eps=config.rms_norm_eps
-        )
-        self.post_attention_layernorm = LlamaRMSNorm(
+        self.input_layernorm = FastRMSNorm.load(prefix=f"{prefix}.input_layernorm", weights=weights, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = FastRMSNorm.load(
             prefix=f"{prefix}.post_attention_layernorm",
             weights=weights,
             eps=config.rms_norm_eps,
@@ -430,7 +353,7 @@ class FlashLlamaModel(torch.nn.Module):
                 for layer_id in range(config.num_hidden_layers)
             ]
         )
-        self.norm = LlamaRMSNorm(
+        self.norm = FastRMSNorm.load(
             prefix="model.norm", weights=weights, eps=config.rms_norm_eps
         )
 

server/text_generation_server/models/custom_modeling/flash_mistral_modeling.py

@@ -35,13 +35,9 @@ from text_generation_server.utils.layers import (
     PositionRotaryEmbedding,
     TensorParallelHead,
     get_linear,
+    FastRMSNorm
 )
-from text_generation_server.utils.import_utils import IS_CUDA_SYSTEM, IS_ROCM_SYSTEM
 
-if IS_CUDA_SYSTEM:
-    import dropout_layer_norm
-elif IS_ROCM_SYSTEM:
-    from vllm import layernorm_ops
 
 if not HAS_FLASH_ATTN_V2_CUDA and not HAS_FLASH_ATTN_V2_ROCM:
     raise ImportError("Mistral model requires flash attn v2")
@@ -100,76 +96,6 @@ class MistralConfig(PretrainedConfig):
             **kwargs,
         )
 
-
-class MistralRMSNorm(nn.Module):
-    def __init__(self, prefix, weights, eps=1e-6):
-        """
-        LlamaRMSNorm is equivalent to T5LayerNorm
-        """
-        super().__init__()
-
-        weight = weights.get_tensor(f"{prefix}.weight")
-        self.weight = nn.Parameter(weight)
-        self.variance_epsilon = eps
-
-    def forward(self, hidden_states, residual=None):
-        if hidden_states.shape[-1] > 8192:
-            if residual is not None:
-                hidden_states += residual
-            residual = hidden_states
-
-            hidden_states = hidden_states.to(torch.float32)
-            variance = hidden_states.pow(2).mean(-1, keepdim=True)
-            hidden_states = hidden_states * torch.rsqrt(
-                variance + self.variance_epsilon
-            )
-
-            # convert into half-precision if necessary
-            if self.weight.dtype in [torch.float16, torch.bfloat16]:
-                hidden_states = hidden_states.to(self.weight.dtype)
-
-            return self.weight * hidden_states, residual
-        elif IS_CUDA_SYSTEM:
-            # faster post attention rms norm
-            normed_hidden_states, res, *rest = dropout_layer_norm.dropout_add_ln_fwd(
-                hidden_states,
-                residual,
-                self.weight,
-                None,
-                None,
-                None,
-                None,
-                None,
-                0.0,
-                self.variance_epsilon,
-                1.0,
-                0,
-                None,
-                False,
-                True,  # Activate RMSNorm
-            )
-            if res is None:
-                res = hidden_states
-
-            return normed_hidden_states, res
-        elif IS_ROCM_SYSTEM:
-            # We use VLLM RMSNorm kernel that can be compiled for RoCm, instead of Flash Attention ones that can not.
-            if residual is not None:
-                hidden_states += residual
-            residual = hidden_states
-
-            out = torch.empty_like(hidden_states)
-            layernorm_ops.rms_norm(
-                out,
-                hidden_states,
-                self.weight.data,
-                self.variance_epsilon,
-            )
-            return out, residual
-        else:
-            raise ValueError("Your system seem to be not supported. Please check your install or open an issue at https://github.com/huggingface/text-generation-inference/issues with a clear reproduction.")
-
-
 def load_attention(config, prefix, weights):
     if config.num_attention_heads != config.num_key_value_heads:
         return _load_gqa(config, prefix, weights)
@@ -371,10 +297,10 @@ class MistralLayer(nn.Module):
         )
         self.mlp = MistralMLP(prefix=f"{prefix}.mlp", config=config, weights=weights)
 
-        self.input_layernorm = MistralRMSNorm(
+        self.input_layernorm = FastRMSNorm.load(
             prefix=f"{prefix}.input_layernorm", weights=weights, eps=config.rms_norm_eps
         )
-        self.post_attention_layernorm = MistralRMSNorm(
+        self.post_attention_layernorm = FastRMSNorm.load(
             prefix=f"{prefix}.post_attention_layernorm",
             weights=weights,
             eps=config.rms_norm_eps,
@@ -440,7 +366,7 @@ class MistralModel(torch.nn.Module):
                 for layer_id in range(config.num_hidden_layers)
             ]
         )
-        self.norm = MistralRMSNorm(
+        self.norm = FastRMSNorm.load(
             prefix="model.norm", weights=weights, eps=config.rms_norm_eps
         )
 

server/text_generation_server/models/custom_modeling/flash_mixtral_modeling.py

@@ -0,0 +1,708 @@
+# coding=utf-8
+# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import torch
+import torch.distributed
+
+import numpy as np
+
+from torch import nn
+from transformers.activations import ACT2FN
+from transformers.configuration_utils import PretrainedConfig
+from typing import Optional, List, Tuple
+
+from text_generation_server.utils import paged_attention, flash_attn
+from text_generation_server.utils.flash_attn import HAS_FLASH_ATTN_V2_ROCM, HAS_FLASH_ATTN_V2_CUDA
+from text_generation_server.utils.layers import (
+    FastLinear,
+    FastRMSNorm,
+    TensorParallelRowLinear,
+    TensorParallelColumnLinear,
+    TensorParallelEmbedding,
+    PositionRotaryEmbedding,
+    TensorParallelHead,
+    get_linear,
+)
+
+if not HAS_FLASH_ATTN_V2_CUDA and not HAS_FLASH_ATTN_V2_ROCM:
+    raise ImportError("Mixtral model requires flash attn v2")
+
+try:
+    import megablocks.ops as ops
+except ImportError:
+    raise ImportError("Mixtral model requires megablocks to be installed")
+
+try:
+    import stk
+except ImportError:
+    raise ImportError("Mixtral model requires stk to be installed")
+
+
+class MixtralConfig(PretrainedConfig):
+    model_type = "mixtral"
+
+    def __init__(
+            self,
+            vocab_size=32000,
+            hidden_size=4096,
+            intermediate_size=14336,
+            num_hidden_layers=32,
+            num_attention_heads=32,
+            num_key_value_heads=8,
+            hidden_act="silu",
+            max_position_embeddings=4096 * 32,
+            initializer_range=0.02,
+            rms_norm_eps=1e-05,
+            use_cache=True,
+            pad_token_id=None,
+            bos_token_id=1,
+            eos_token_id=2,
+            pretraining_tp=1,
+            tie_word_embeddings=False,
+            rope_theta=10000.0,
+            sliding_window=4096,
+            num_experts_per_tok=2,
+            num_local_experts=8,
+            **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.sliding_window = sliding_window
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.pretraining_tp = pretraining_tp
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.num_experts_per_tok = num_experts_per_tok
+        self.num_local_experts = num_local_experts
+
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+
+def promote_scalar(x: torch.Tensor) -> torch.Tensor:
+    return x.view(1) if len(x.size()) == 0 else x
+
+
+def load_attention(config, prefix, weights):
+    if config.num_attention_heads != config.num_key_value_heads:
+        return _load_gqa(config, prefix, weights)
+    else:
+        return TensorParallelColumnLinear.load_multi(
+            config,
+            prefixes=[f"{prefix}.q_proj", f"{prefix}.k_proj", f"{prefix}.v_proj"],
+            dim=0,
+            weights=weights,
+            bias=False,
+        )
+
+
+def _load_gqa(config, prefix: str, weights):
+    assert config.hidden_size % config.num_attention_heads == 0
+    assert config.num_attention_heads % weights.process_group.size() == 0
+
+    weight = weights.get_multi_weights_col(
+        prefixes=[f"{prefix}.q_proj", f"{prefix}.k_proj", f"{prefix}.v_proj"],
+        quantize=config.quantize,
+        dim=0,
+    )
+
+    if config.quantize not in ["gptq", "awq"]:
+        weight = weight.to(dtype=weights.dtype).to(device=weights.device)
+
+        head_size = config.hidden_size // config.num_attention_heads
+        num_heads = config.num_attention_heads // weights.process_group.size()
+        num_key_value_heads = config.num_key_value_heads // weights.process_group.size()
+        assert list(weight.shape) == [
+            (num_heads + 2 * num_key_value_heads) * head_size,
+            config.hidden_size,
+        ], f"{list(weight.shape)} != {[(num_heads + 2 * config.num_key_value_heads) * head_size, config.hidden_size]}"
+
+    return TensorParallelColumnLinear(
+        get_linear(weight, bias=None, quantize=config.quantize)
+    )
+
+
+def _load_experts(config, prefix, mat, weights):
+    if config.quantize is not None:
+        raise NotImplementedError("Mixtral does not support weight quantization yet.")
+
+    assert mat in ["w1", "w2", "w3"]
+
+    world_size = weights.process_group.size()
+    rank = weights.process_group.rank()
+
+    assert (
+            config.intermediate_size % world_size == 0
+    ), f"The chosen size {config.intermediate_size} is not compatible with sharding on {world_size} shards"
+
+    block_size = config.intermediate_size // world_size
+    start = rank * block_size
+    stop = (rank + 1) * block_size
+
+    tensor = torch.empty((config.num_local_experts * block_size, config.hidden_size),
+                         dtype=weights.dtype,
+                         device=weights.device)
+
+    for i in range(config.num_local_experts):
+        slice_ = weights._get_slice(f"{prefix}.{i}.{mat}.weight")
+
+        if mat == "w2":
+            expert_slice = slice_[:, start:stop].t().contiguous()
+        else:
+            expert_slice = slice_[start:stop]
+        tensor[i * block_size:(i + 1) * block_size] = expert_slice.to(dtype=weights.dtype).to(device=weights.device)
+    return tensor
+
+
+class MixtralAttention(torch.nn.Module):
+    def __init__(
+            self,
+            prefix: str,
+            config,
+            weights,
+    ):
+        super().__init__()
+        self.max_past = (
+            config.sliding_window if config.sliding_window is not None else 0
+        )
+        self.num_heads = config.num_attention_heads
+        self.hidden_size = config.hidden_size
+        self.head_size = self.hidden_size // self.num_heads
+
+        self.rotary_emb = PositionRotaryEmbedding.static(
+            config=config,
+            dim=self.head_size,
+            base=config.rope_theta,
+            device=weights.device,
+        )
+
+        self.softmax_scale = self.head_size ** -0.5
+
+        if self.num_heads % weights.process_group.size() != 0:
+            raise ValueError(
+                f"`num_heads` must be divisible by `num_shards` (got `num_heads`: {self.num_heads} "
+                f"and `num_shards`: {weights.process_group.size()}"
+            )
+        self.num_heads = self.num_heads // weights.process_group.size()
+        self.num_key_value_heads = (
+                config.num_key_value_heads // weights.process_group.size()
+        )
+
+        self.query_key_value = load_attention(config, prefix, weights)
+
+        self.o_proj = TensorParallelRowLinear.load(
+            config,
+            prefix=f"{prefix}.o_proj",
+            weights=weights,
+            bias=False,
+        )
+        self.num_groups = self.num_heads // self.num_key_value_heads
+        self.kv_head_mapping = torch.arange(
+            0, self.num_key_value_heads, dtype=torch.int32, device=weights.device
+        ).repeat_interleave(self.num_groups)
+
+    def forward(
+            self,
+            hidden_states,
+            cos,
+            sin,
+            cu_seqlen_prefill,
+            kv_cache,
+            block_tables,
+            slots,
+            input_lengths,
+            max_s,
+            prefill_cache_indices,
+    ):
+        qkv = self.query_key_value(hidden_states)
+        query, kv = qkv.split(
+            [
+                self.head_size * self.num_heads,
+                2 * self.head_size * self.num_key_value_heads,
+            ],
+            dim=1,
+        )
+        query = query.view(-1, self.num_heads, self.head_size)
+        kv = kv.view(-1, 2, self.num_key_value_heads, self.head_size)
+
+        self.rotary_emb(query, torch.select(kv, dim=1, index=0), cos, sin)
+
+        if prefill_cache_indices is not None:
+            kv_to_cache = kv[prefill_cache_indices]
+        else:
+            kv_to_cache = kv
+
+        paged_attention.reshape_and_cache(
+            kv_to_cache[:, 0], kv_to_cache[:, 1], kv_cache[0], kv_cache[1], slots
+        )
+
+        # output tensor
+        attn_output = torch.empty_like(query)
+
+        # Prefill
+        if cu_seqlen_prefill is not None:
+            # flash attention
+            flash_attn.attention(
+                query,
+                torch.select(kv, dim=1, index=0),
+                torch.select(kv, dim=1, index=1),
+                attn_output,
+                cu_seqlen_prefill,
+                max_s,
+                self.softmax_scale,
+                window_size_left=self.max_past,
+            )
+        # Decode
+        else:
+            paged_attention.attention(
+                attn_output,
+                query,
+                kv_cache[0],
+                kv_cache[1],
+                self.kv_head_mapping,
+                self.softmax_scale,
+                block_tables,
+                input_lengths,
+                max_s,
+            )
+
+        return self.o_proj(attn_output.view(-1, self.num_heads * self.head_size))
+
+
+@torch.jit.script
+def select_experts(gate_logits: torch.Tensor, top_k: int):
+    # all_probs: (sequence_length, n_experts) and upcast for softmax
+    all_probs = torch.nn.functional.softmax(gate_logits, dim=1, dtype=torch.float)
+    # weights, selected_experts: (sequence_length, top-k)
+    weights, selected_experts = torch.topk(all_probs, top_k, dim=-1)
+    weights /= weights.sum(dim=-1, keepdim=True)
+    weights = weights.view(-1)
+    selected_experts = selected_experts.view(-1)
+
+    return selected_experts, weights
+
+
+@torch.jit.script
+def round_up(x: torch.Tensor, value: int):
+    return torch.div(x + (value - 1), value, rounding_mode="trunc") * value
+
+
+class BlockSparseMoE(nn.Module):
+    """
+    Built on the paper and library Megablocks as described in
+    https://arxiv.org/abs/2211.15841. This implementation is
+    strictly equivalent to standard MoE with full capacity (no
+    dropped tokens). It's faster since it formulates MoE operations
+    in terms of block-sparse operations to accomodate imbalanced
+    assignments of tokens to experts, whereas standard MoE either
+    (1) drop tokens at the cost of reduced performance or (2) set
+    capacity factor to number of experts and thus waste computation
+    and memory on padding.
+    """
+
+    def __init__(self, prefix, config: MixtralConfig, weights):
+        super().__init__()
+        self.hidden_dim = config.hidden_size
+        self.ffn_dim = config.intermediate_size // weights.process_group.size()
+        self.num_experts = config.num_local_experts
+        self.top_k = config.num_experts_per_tok
+
+        act = config.hidden_act
+        if "gelu" in act:
+            self.act = lambda x: torch.nn.functional.gelu(
+                x,
+                approximate="tanh"
+                if act in ["gelu_fast", "gelu_pytorch_tanh"]
+                else "none",
+            )
+        elif "silu" in act:
+            self.act = torch.nn.functional.silu
+        else:
+            self.act = ACT2FN[act]
+
+        # gating
+        self.gate = FastLinear.load(config, f"{prefix}.gate", weights, bias=False)
+
+        # merged expert weights, all of size  (n_experts * ffn_dim, hidden_dim)
+        self.w1 = _load_experts(config, f"{prefix}.experts", "w1", weights).t()
+        self.w2 = _load_experts(config, f"{prefix}.experts", "w2", weights)
+        self.w3 = _load_experts(config, f"{prefix}.experts", "w3", weights).t()
+
+        self.offsets = None
+        self.offsets_block_rows = 0
+
+        self.process_group = weights.process_group
+
+        # Calculate the number of bits needed to represent the expert indices
+        # so that we can pass it to radix sort.
+        self.sort_end_bit = max(int(np.ceil(np.log2(self.num_experts))), 1)
+        self.blocking = 128
+        self.quantize_scatter_num_bits = -1
+
+    def topology(self, x: torch.Tensor, padded_bins: torch.Tensor):
+        padded_tokens, _ = x.size()
+        assert padded_tokens % self.blocking == 0
+        assert self.ffn_dim % self.blocking == 0
+
+        # Offsets for the sparse matrix. All rows have the
+        # same number of nonzero blocks dictated by the
+        # dimensionality of a single expert.
+        block_rows = padded_tokens // self.blocking
+        blocks_per_row = self.ffn_dim // self.blocking
+        if self.offsets is None or block_rows > self.offsets_block_rows:
+            self.offsets = torch.arange(
+                0,
+                block_rows * blocks_per_row + 1,
+                blocks_per_row,
+                dtype=torch.int32,
+                device=x.device,
+            )
+            self.offsets_block_rows = block_rows
+            offsets = self.offsets
+        else:
+            offsets = self.offsets[:block_rows]
+
+        # Indices for the sparse matrix. The indices for
+        # the intermediate matrix are dynamic depending
+        # on the mapping of tokens to experts.
+        column_indices = ops.topology(padded_bins, self.blocking, block_rows,
+                                      blocks_per_row)
+
+        # For now, use meta init to save the device memory.
+        data = torch.empty(
+            column_indices.numel(),
+            self.blocking,
+            self.blocking,
+            dtype=x.dtype,
+            device="meta",
+        )
+        shape = (padded_tokens, self.ffn_dim * self.num_experts)
+        row_indices = stk.ops.row_indices(shape, data, offsets, column_indices)
+        return stk.Matrix(
+            shape,
+            data,
+            row_indices,
+            column_indices,
+            offsets,
+            False,
+            False,
+            False,
+        )
+
+    def indices_and_padded_bins(self, selected_experts: torch.Tensor):
+        # Sort the expert ids to produce the scatter/gather
+        # indices for the permutation.
+        # selected_experts = selected_experts.int()
+
+        # returns bin_ids == num of experts for this sequence ? == unique selected experts?
+        # and indices == how to sort tokens?
+        bin_ids, indices = ops.sort(selected_experts, self.sort_end_bit)
+        # bin_ids => [0, 0, 0, 2, 2, ...] => [num_tokens * top_k]
+        # indices => [14, 32, 33, ...] => [num_tokens * top_k]
+
+        # Histogram the expert ids to identify the number of
+        # tokens routed to each expert.
+        tokens_per_expert = ops.histogram(selected_experts, self.num_experts)
+        # tokens_per_expert => [3, 0, 2, ...] => [num_experts]
+
+        # Round the token counts up to the block size used in
+        # the matrix muliplications. Caculate the starting
+        # position of each bin.
+
+        # List of size num_experts
+        padded_tokens_per_expert = round_up(tokens_per_expert,
+                                            self.blocking)
+        # padded_tokens_per_expert => [128, O, 128, ...]
+
+        # Cumulative selected experts per token
+        padded_bins = ops.inclusive_cumsum(padded_tokens_per_expert, 0)
+        padded_bins = promote_scalar(padded_bins)
+        # padded_bins => [128, 128, 256, ...]
+
+        # Calculate the bin bounds for the sorted tokens.
+        bins = ops.inclusive_cumsum(tokens_per_expert, 0)
+        bins = promote_scalar(bins)
+        # bins => [3, 3, 5, ...]
+
+        return indices, bin_ids, bins, padded_bins, tokens_per_expert
+
+    @torch.inference_mode()
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        x: (sequence_length, model_dim)
+        gate_logits: (sequence_length, n_experts)
+        """
+        # optional reshape
+        input_shape = x.shape
+        x = x.view(-1, input_shape[-1])
+
+        # gate_logits: (sequence_length, n_experts)
+        gate_logits = self.gate(x)
+        selected_experts, weights = select_experts(gate_logits, self.top_k)
+
+        (
+            indices,
+            bin_ids,
+            bins,
+            padded_bins,
+            _,
+        ) = self.indices_and_padded_bins(selected_experts)
+
+        # Permute tokens and pad to prepare expert computation
+        # (top_k * sequence_length + padding, model_dim)
+        x = ops.padded_gather(x, indices, bin_ids, bins, padded_bins,
+                              self.top_k)
+
+        # Create the sparse matrix topology
+        with torch.no_grad():
+            topo = self.topology(x, padded_bins)
+
+        # Perform the expert computation
+        # First Dense x Dense -> Sparse for w1 and w3,
+        # (top_k * sequence_length + padding, ffn_dim * n_experts)
+        x = stk.Matrix(
+            topo.size(),
+            self.act(stk.ops.sdd(x, self.w1, topo).data) *
+            stk.ops.sdd(x, self.w3, topo).data,
+            topo.row_indices,
+            topo.column_indices,
+            topo.offsets,
+            topo.column_indices_t,
+            topo.offsets_t,
+            topo.block_offsets_t,
+        )
+
+        # Then Sparse x Dense -> Dense for w2
+        # (top_k * sequence_length + padding, model_dim)
+        x = stk.ops.dsd(x, self.w2)
+
+        # Permute back and remove padding
+        # (sequence_length, model_dim)
+        x = ops.padded_scatter(
+            x,
+            indices,
+            bin_ids,
+            weights,
+            bins,
+            padded_bins,
+            self.top_k,
+            self.quantize_scatter_num_bits,
+        ).view(*input_shape)
+
+        if self.process_group.size() > 1:
+            torch.distributed.all_reduce(x, group=self.process_group)
+
+        return x.view(*input_shape)
+
+
+class MixtralLayer(nn.Module):
+    def __init__(self, layer_id, config, weights):
+        super().__init__()
+        prefix = f"model.layers.{layer_id}"
+
+        self.self_attn = MixtralAttention(
+            prefix=f"{prefix}.self_attn", config=config, weights=weights
+        )
+        self.block_sparse_moe = BlockSparseMoE(f"{prefix}.block_sparse_moe", config, weights)
+
+        self.input_layernorm = FastRMSNorm.load(
+            prefix=f"{prefix}.input_layernorm", weights=weights, eps=config.rms_norm_eps
+        )
+        self.post_attention_layernorm = FastRMSNorm.load(
+            prefix=f"{prefix}.post_attention_layernorm",
+            weights=weights,
+            eps=config.rms_norm_eps,
+        )
+
+    def forward(
+            self,
+            hidden_states,
+            residual,
+            cos,
+            sin,
+            cu_seqlen_prefill,
+            kv_cache,
+            block_tables,
+            slots,
+            input_lengths,
+            max_s,
+            prefill_cache_indices,
+    ):
+        normed_hidden_states, res = self.input_layernorm(hidden_states, residual)
+
+        # Self Attention
+        attn_output = self.self_attn(
+            normed_hidden_states,
+            cos,
+            sin,
+            cu_seqlen_prefill,
+            kv_cache,
+            block_tables,
+            slots,
+            input_lengths,
+            max_s,
+            prefill_cache_indices,
+        )
+
+        # faster post attention rms norm
+        normed_attn_res_output, attn_res = self.post_attention_layernorm(
+            attn_output, res
+        )
+
+        block_sparse_moe_output = self.block_sparse_moe(normed_attn_res_output)
+
+        return block_sparse_moe_output, attn_res
+
+
+class MixtralModel(torch.nn.Module):
+    def __init__(self, config, weights):
+        super().__init__()
+
+        self.embed_tokens = TensorParallelEmbedding(
+            prefix="model.embed_tokens", weights=weights
+        )
+
+        self.layers = nn.ModuleList(
+            [
+                MixtralLayer(
+                    layer_id,
+                    config,
+                    weights,
+                )
+                for layer_id in range(config.num_hidden_layers)
+            ]
+        )
+        self.norm = FastRMSNorm.load(
+            prefix="model.norm", weights=weights, eps=config.rms_norm_eps
+        )
+
+        self.head_size = self.layers[0].self_attn.head_size
+        self.num_heads = self.layers[0].self_attn.num_heads
+        self.num_key_value_heads = self.layers[0].self_attn.num_key_value_heads
+
+    def forward(
+            self,
+            input_ids: torch.Tensor,
+            position_ids: torch.Tensor,
+            cu_seqlen_prefill: Optional[torch.Tensor],
+            kv_cache: List[Tuple[torch.Tensor, torch.Tensor]],
+            block_tables: torch.Tensor,
+            slots: torch.Tensor,
+            input_lengths: torch.Tensor,
+            max_s: int,
+            prefill_cache_indices: Optional[torch.Tensor],
+    ) -> torch.Tensor:
+        hidden_states = self.embed_tokens(input_ids)
+
+        # Get rotary cos and sin for this forward
+        # Avoid to index in each layer
+        cos, sin = self.layers[0].self_attn.rotary_emb.get_cos_sin(
+            position_ids, max_s, hidden_states.dtype
+        )
+
+        residual = None
+        for i, layer in enumerate(self.layers):
+            hidden_states, residual = layer(
+                hidden_states,
+                residual,
+                cos,
+                sin,
+                cu_seqlen_prefill,
+                kv_cache[i],
+                block_tables,
+                slots,
+                input_lengths,
+                max_s,
+                prefill_cache_indices,
+            )
+
+        hidden_states, _ = self.norm(hidden_states, residual)
+
+        return hidden_states
+
+
+class FlashMixtralForCausalLM(torch.nn.Module):
+    def __init__(self, config, weights):
+        super().__init__()
+
+        self.model = MixtralModel(config, weights)
+        self.lm_head = TensorParallelHead.load(
+            config,
+            prefix="lm_head",
+            weights=weights,
+        )
+        self.max_past = config.sliding_window
+        if self.max_past is None:
+            raise ValueError("max_past cannot be None")
+
+    def forward(
+            self,
+            input_ids: torch.Tensor,
+            position_ids: torch.Tensor,
+            cu_seqlen_prefill: Optional[torch.Tensor],
+            kv_cache: List[Tuple[torch.Tensor, torch.Tensor]],
+            block_tables: torch.Tensor,
+            slots: torch.Tensor,
+            input_lengths: torch.Tensor,
+            max_s: int,
+            prefill_cache_indices: Optional[torch.Tensor],
+            lm_head_indices: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        if prefill_cache_indices is not None:
+            # Slots also need to be sliced as it has the same size as the whole kv tensor
+            slots = slots[prefill_cache_indices]
+        else:
+            # Clamp in decode mode as paged attention requires clamped values whereas the flash attention
+            # kernel requires the true values
+            max_s = min(self.max_past, max_s)
+            input_lengths = torch.clamp(input_lengths, max=self.max_past)
+
+        hidden_states = self.model(
+            input_ids,
+            position_ids,
+            cu_seqlen_prefill,
+            kv_cache,
+            block_tables,
+            slots,
+            input_lengths,
+            max_s,
+            prefill_cache_indices,
+        )
+        if lm_head_indices is not None:
+            hidden_states = hidden_states[lm_head_indices]
+        logits = self.lm_head(hidden_states)
+        return logits

server/text_generation_server/models/custom_modeling/flash_santacoder_modeling.py

@@ -6,7 +6,6 @@ from transformers.activations import ACT2FN
 from typing import Optional, List, Tuple
 
 from text_generation_server.utils import paged_attention, flash_attn
-from text_generation_server.utils.flash_attn import attention
 from text_generation_server.utils.layers import (
     TensorParallelRowLinear,
     TensorParallelColumnLinear,

server/text_generation_server/models/flash_mistral.py

@@ -8,14 +8,13 @@ from dataclasses import dataclass
 from opentelemetry import trace
 from transformers import PreTrainedTokenizerBase
 from transformers.models.llama import LlamaTokenizerFast
-from typing import Optional, Tuple, Type
+from typing import Optional, Tuple, Type, List
 
 from text_generation_server.pb import generate_pb2
 from text_generation_server.models import FlashCausalLM
 from text_generation_server.models.flash_causal_lm import FlashCausalLMBatch, BLOCK_SIZE
 from text_generation_server.models.cache_manager import (
     get_cache_manager,
-    set_cache_manager,
 )
 from text_generation_server.models.custom_modeling.flash_mistral_modeling import (
     FlashMistralForCausalLM,
@@ -278,9 +277,11 @@ class FlashMistralBatch(FlashCausalLMBatch):
         )
 
 
-class FlashMistral(FlashCausalLM):
+class BaseFlashMistral(FlashCausalLM):
     def __init__(
             self,
+            config_cls,
+            model_cls,
             model_id: str,
             revision: Optional[str] = None,
             quantize: Optional[str] = None,
@@ -305,7 +306,7 @@ class FlashMistral(FlashCausalLM):
             trust_remote_code=trust_remote_code,
         )
 
-        config = MistralConfig.from_pretrained(
+        config = config_cls.from_pretrained(
             model_id, revision=revision, trust_remote_code=trust_remote_code
         )
         config.quantize = quantize
@@ -321,10 +322,10 @@ class FlashMistral(FlashCausalLM):
         if config.quantize in ["gptq", "awq"]:
             weights._set_gptq_params(model_id)
 
-        model = FlashMistralForCausalLM(config, weights)
+        model = model_cls(config, weights)
 
         torch.distributed.barrier(group=self.process_group)
-        super(FlashMistral, self).__init__(
+        super(BaseFlashMistral, self).__init__(
             model=model,
             tokenizer=tokenizer,
             num_layers=len(model.model.layers),
@@ -396,3 +397,23 @@ class FlashMistral(FlashCausalLM):
         if batch.prefill_cache_indices is not None:
             batch.prefill_cache_indices = None
         return logits
+
+
+class FlashMistral(BaseFlashMistral):
+    def __init__(
+            self,
+            model_id: str,
+            revision: Optional[str] = None,
+            quantize: Optional[str] = None,
+            dtype: Optional[torch.dtype] = None,
+            trust_remote_code: bool = False,
+    ):
+        super(FlashMistral, self).__init__(
+            config_cls=MistralConfig,
+            model_cls=FlashMistralForCausalLM,
+            model_id=model_id,
+            revision=revision,
+            quantize=quantize,
+            dtype=dtype,
+            trust_remote_code=trust_remote_code
+        )

server/text_generation_server/models/flash_mixtral.py

@@ -0,0 +1,26 @@
+import torch
+
+from typing import Optional
+
+from text_generation_server.models.flash_mistral import BaseFlashMistral
+from text_generation_server.models.custom_modeling.flash_mixtral_modeling import MixtralConfig, FlashMixtralForCausalLM
+
+
+class FlashMixtral(BaseFlashMistral):
+    def __init__(
+            self,
+            model_id: str,
+            revision: Optional[str] = None,
+            quantize: Optional[str] = None,
+            dtype: Optional[torch.dtype] = None,
+            trust_remote_code: bool = False,
+    ):
+        super(FlashMixtral, self).__init__(
+            config_cls=MixtralConfig,
+            model_cls=FlashMixtralForCausalLM,
+            model_id=model_id,
+            revision=revision,
+            quantize=quantize,
+            dtype=dtype,
+            trust_remote_code=trust_remote_code
+        )

server/text_generation_server/utils/layers.py

@@ -47,12 +47,14 @@ elif CAN_EXLLAMA:
                                                                      create_exllama_buffers,
                                                                      set_device,
                                                                      )
+
             HAS_EXLLAMA = "2"
         else:
             from text_generation_server.utils.gptq.exllama import (Ex4bitLinear as ExllamaQuantLinear,
                                                                    create_exllama_buffers,
                                                                    set_device,
                                                                    )
+
             HAS_EXLLAMA = "1"
 
     except ImportError:
@@ -526,9 +528,12 @@ class TensorParallelEmbedding(nn.Module):
 try:
     if IS_CUDA_SYSTEM:
         import dropout_layer_norm
+    elif IS_ROCM_SYSTEM:
+        from vllm import layernorm_ops
     else:
         dropout_layer_norm = None
 
+
     class FastLayerNorm(nn.LayerNorm):
         def forward(self, hidden_states, residual=None):
             if hidden_states.shape[-1] > 8192 or IS_ROCM_SYSTEM:
@@ -563,10 +568,81 @@ try:
                     residual = hidden_states
 
                 return normed_hidden_states, residual
+
+
+    class FastRMSNorm(nn.Module):
+        def __init__(self, weight: torch.Tensor, eps: float):
+            super().__init__()
+
+            self.weight = nn.Parameter(weight)
+            self.variance_epsilon = eps
+
+        @classmethod
+        def load(cls, prefix, weights, eps=1e-6):
+            weight = weights.get_tensor(f"{prefix}.weight")
+            return cls(weight, eps)
+
+        def forward(self, hidden_states, residual=None):
+            if hidden_states.shape[-1] > 8192:
+                if residual is not None:
+                    hidden_states += residual
+                residual = hidden_states
+
+                hidden_states = hidden_states.to(torch.float32)
+                variance = hidden_states.pow(2).mean(-1, keepdim=True)
+                hidden_states = hidden_states * torch.rsqrt(
+                    variance + self.variance_epsilon
+                )
+
+                # convert into half-precision if necessary
+                if self.weight.dtype in [torch.float16, torch.bfloat16]:
+                    hidden_states = hidden_states.to(self.weight.dtype)
+
+                return self.weight * hidden_states, residual
+            elif IS_CUDA_SYSTEM:
+                # faster post attention rms norm
+                normed_hidden_states, res, *rest = dropout_layer_norm.dropout_add_ln_fwd(
+                    hidden_states,
+                    residual,
+                    self.weight,
+                    None,
+                    None,
+                    None,
+                    None,
+                    None,
+                    0.0,
+                    self.variance_epsilon,
+                    1.0,
+                    0,
+                    None,
+                    False,
+                    True,  # Activate RMSNorm
+                )
+                if res is None:
+                    res = hidden_states
+
+                return normed_hidden_states, res
+            elif IS_ROCM_SYSTEM:
+                # We use VLLM RMSNorm kernel that can be compiled for RoCm, instead of Flash Attention ones that can not.
+                if residual is not None:
+                    hidden_states += residual
+                residual = hidden_states
+
+                out = torch.empty_like(hidden_states)
+                layernorm_ops.rms_norm(
+                    out,
+                    hidden_states,
+                    self.weight.data,
+                    self.variance_epsilon,
+                )
+                return out, residual
+            else:
+                raise ValueError(
+                    "Your system seem to be not supported. Please check your install or open an issue at https://github.com/huggingface/text-generation-inference/issues with a clear reproduction.")
+
 except ImportError:
     pass
 
-
 try:
     if IS_CUDA_SYSTEM:
         from flash_attn.layers.rotary import RotaryEmbedding
@@ -574,12 +650,14 @@ try:
     elif IS_ROCM_SYSTEM:
         from vllm import pos_encoding_ops
 
+
     def _create_inv_freq(dim, base, device):
         inv_freq = 1.0 / (
                 base ** (torch.arange(0, dim, 2, device=device, dtype=torch.float32) / dim)
         )
         return inv_freq
 
+
     def _get_rope_config(config):
         if os.getenv("ROPE_SCALING", None) is not None:
             rope_scaling = {
@@ -589,6 +667,7 @@ try:
             return rope_scaling
         return getattr(config, "rope_scaling", None)
 
+
     class PositionRotaryEmbedding(nn.Module):
         def __init__(self, inv_freq, scaling_factor):
             super().__init__()
@@ -630,7 +709,8 @@ try:
                     True
                 )
             else:
-                raise ValueError("Your system seem to be not supported. Please check your install or open an issue at https://github.com/huggingface/text-generation-inference/issues with a clear reproduction.")
+                raise ValueError(
+                    "Your system seem to be not supported. Please check your install or open an issue at https://github.com/huggingface/text-generation-inference/issues with a clear reproduction.")
 
         @classmethod
         def static(cls, config, dim, base, device):
@@ -747,6 +827,7 @@ try:
             # Note: this unsqueeze is not necessary on RoCm + VLLM ROPE implementation, but we leave it as is to avoid yet an other controlflow.
             return cos.unsqueeze(1), sin.unsqueeze(1)
 
+
     class DynamicPositionRotaryEmbedding(PositionRotaryEmbedding):
         def __init__(self, dim, max_position_embeddings, base, device, scaling_factor):
             inv_freq = _create_inv_freq(dim, base, device)
@@ -783,9 +864,12 @@ try:
 
     # Inverse dim formula to find dim based on number of rotations
     import math
+
+
     def find_correction_dim(num_rotations, dim, base=10000, max_position_embeddings=2048):
         return (dim * math.log(max_position_embeddings / (num_rotations * 2 * math.pi))) / (2 * math.log(base))
 
+
     # Find dim range bounds based on rotations
     def find_correction_range(low_rot, high_rot, dim, base=10000, max_position_embeddings=2048):
         low = math.floor(find_correction_dim(
@@ -794,6 +878,7 @@ try:
             high_rot, dim, base, max_position_embeddings))
         return max(low, 0), min(high, dim - 1)  # Clamp values just in case
 
+
     def linear_ramp_mask(min, max, dim):
         if min == max:
             max += 0.001  # Prevent singularity
@@ -802,13 +887,16 @@ try:
         ramp_func = torch.clamp(linear_func, 0, 1)
         return ramp_func
 
+
     def get_mscale(scale=1):
         if scale <= 1:
             return 1.0
         return 0.1 * math.log(scale) + 1.0
 
+
     class YarnPositionRotaryEmbedding(PositionRotaryEmbedding):
-        def __init__(self, dim, max_position_embeddings, base, device, scaling_factor,*, extrapolation_factor, attn_factor, beta_fast, beta_slow):
+        def __init__(self, dim, max_position_embeddings, base, device, scaling_factor, *, extrapolation_factor,
+                     attn_factor, beta_fast, beta_slow):
             inv_freq = _create_inv_freq(dim, base, device)
             super().__init__(inv_freq, scaling_factor)
             self.dim = dim
@@ -818,7 +906,8 @@ try:
             self.attn_factor = attn_factor
             self.beta_fast = beta_fast
             self.beta_slow = beta_slow
-            self.mscale = float(get_mscale(self.scaling_factor) * self.attn_factor) # Get n-d magnitude scaling corrected for interpolation
+            self.mscale = float(get_mscale(
+                self.scaling_factor) * self.attn_factor)  # Get n-d magnitude scaling corrected for interpolation
 
         def _update_cos_sin_cache(self, dtype, device, seqlen):
             # Reset the tables if the sequence length has changed,
@@ -834,13 +923,15 @@ try:
                     )
                     freqs = 1.0 / inv_freq_extrapolation
                     inv_freq_interpolation = 1.0 / (self.scaling_factor * freqs)
-                    low, high = find_correction_range(self.beta_fast, self.beta_slow, self.dim, self.base, self.max_position_embeddings)
-                    inv_freq_mask = (1 - linear_ramp_mask(low, high, self.dim // 2).float().to(device)) * self.extrapolation_factor # Get n-d rotational scaling corrected for extrapolation
+                    low, high = find_correction_range(self.beta_fast, self.beta_slow, self.dim, self.base,
+                                                      self.max_position_embeddings)
+                    inv_freq_mask = (1 - linear_ramp_mask(low, high, self.dim // 2).float().to(
+                        device)) * self.extrapolation_factor  # Get n-d rotational scaling corrected for extrapolation
                     inv_freq = inv_freq_interpolation * (1 - inv_freq_mask) + inv_freq_extrapolation * inv_freq_mask
 
                     self.inv_freq = inv_freq
-                    self.mscale = float(get_mscale(self.scaling_factor) * self.attn_factor) # Get n-d magnitude scaling corrected for interpolation
-
+                    self.mscale = float(get_mscale(
+                        self.scaling_factor) * self.attn_factor)  # Get n-d magnitude scaling corrected for interpolation
 
                 self._seq_len_cached = seqlen
                 t = torch.arange(seqlen, device=device, dtype=self.inv_freq.dtype)