WIP, many bits are still missing...

So this won't work yet.

server/text_generation_server/layers/tensor_parallel.py

@@ -146,7 +146,11 @@ class TensorParallelColumnLinear(SuperLayer):
             num_key_value_heads=num_key_value_heads,
         )
         if bias:
-            raise NotImplementedError("packed_qkv only implemented for baichuan")
+            bias = weights.get_packed_sharded(
+                f"{prefix}.bias",
+                dim=0,
+                block_sizes=[num_heads, num_key_value_heads, num_key_value_heads],
+            )
         else:
             bias = None
         linear = get_linear(weight, bias, config.quantize)

server/text_generation_server/models/__init__.py

@@ -82,12 +82,14 @@ try:
     from text_generation_server.models.flash_phi import FlashPhi
     from text_generation_server.models.flash_starcoder2 import FlashStarcoder2
     from text_generation_server.models.flash_dbrx import FlashDbrx
+    from text_generation_server.models.flash_phi3small import FlashPhi3Small
     from text_generation_server.layers.attention import SUPPORTS_WINDOWING
 except ImportError as e:
     logger.warning(f"Could not import Flash Attention enabled models: {e}")
     SUPPORTS_WINDOWING = False
     FLASH_ATTENTION = False
 
+
 if FLASH_ATTENTION:
     __all__.append(FlashGPT2)
     __all__.append(FlashNeoXSharded)
@@ -103,6 +105,7 @@ if FLASH_ATTENTION:
     __all__.append(FlashStarcoder2)
     __all__.append(FlashGemma)
     __all__.append(FlashCohere)
+    __all__.append(FlashPhi3Small)
 
 MAMBA_AVAILABLE = True
 try:
@@ -138,6 +141,11 @@ class ModelType(enum.Enum):
         "name": "Phi 3",
         "url": "https://huggingface.co/microsoft/Phi-3-mini-4k-instruct",
     }
+    PHI3_SMALL = {
+        "type": "phi3small",
+        "name": "Phi 3 Small",
+        "url": "https://huggingface.co/microsoft/Phi-3-small-8k-instruct",
+    }
     GEMMA = {
         "type": "gemma",
         "name": "Gemma",
@@ -864,6 +872,30 @@ def get_model(
         else:
             raise NotImplementedError(FLASH_ATT_ERROR_MESSAGE.format("LlavaNext"))
 
+    if model_type == PHI3_SMALL:
+        if FLASH_ATTENTION:
+            return FlashPhi3Small(
+                model_id,
+                revision,
+                quantize=quantize,
+                speculator=speculator,
+                dtype=dtype,
+                trust_remote_code=trust_remote_code,
+            )
+        elif sharded:
+            raise NotImplementedError(
+                FLASH_ATT_ERROR_MESSAGE.format("Sharded Phi3 Small")
+            )
+        else:
+            return CausalLM(
+                model_id,
+                revision,
+                quantize=quantize,
+                speculator=speculator,
+                dtype=dtype,
+                trust_remote_code=trust_remote_code,
+            )
+
     if sharded:
         raise NotImplementedError("sharded is not supported for AutoModel")
     if quantize == "gptq":

server/text_generation_server/models/custom_modeling/flash_phi3small_modeling.py

@@ -0,0 +1,434 @@
+# 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.
+
+from typing import List, Optional, Tuple
+
+import torch
+import torch.distributed
+
+from torch import nn
+from transformers.activations import ACT2FN
+
+from text_generation_server.utils.import_utils import SYSTEM
+from text_generation_server.layers.attention import (
+    paged_attention,
+    attention,
+    reshape_and_cache,
+)
+from text_generation_server.layers import (
+    TensorParallelRowLinear,
+    TensorParallelColumnLinear,
+    TensorParallelEmbedding,
+    SpeculativeHead,
+)
+from text_generation_server.layers.rotary import PositionRotaryEmbedding
+from text_generation_server.layers.layernorm import (
+    FastLayerNorm,
+)
+
+if SYSTEM == "rocm":
+    try:
+        from vllm import _custom_C
+    except Exception as e:
+        raise ImportError(f"Could not load `vllm._custom_C`. Full error: {e}")
+
+
+def load_attention(config, prefix, weights):
+    # Only defined in granite.
+    bias = getattr(config, "attention_bias", False)
+    bias = True
+    return TensorParallelColumnLinear.load_qkv(
+        config,
+        prefix=f"{prefix}.query_key_value",
+        weights=weights,
+        bias=bias,
+        num_heads=config.num_attention_heads,
+        num_key_value_heads=config.num_key_value_heads,
+    )
+
+
+class FlashPhi3SmallAttention(torch.nn.Module):
+    def __init__(
+        self,
+        prefix: str,
+        config,
+        weights,
+    ):
+        super().__init__()
+        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_embedding_base,
+            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()}"
+            )
+        if config.num_key_value_heads % weights.process_group.size() != 0:
+            raise ValueError(
+                f"`num_key_value_heads` must be divisible by `num_shards` (got `num_key_value_heads`: {config.num_key_value_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}.dense",
+            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,
+    ):
+        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)
+
+        reshape_and_cache(kv[:, 0], kv[:, 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
+            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,
+            )
+        # Decode
+        else:
+            paged_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))
+
+
+# Copyright (c) Microsoft Corporation, MIT license
+# https://huggingface.co/microsoft/Phi-3-small-8k-instruct/blob/e6adf2a152c4cede5e9b88ede22ce5c0af2861fc/modeling_phi3_small.py#L88
+# Remove once this is in transformers.
+def gegelu(input: torch.Tensor, limit: Optional[float] = None):
+    a_gelu, a_linear = input[..., ::2], input[..., 1::2]
+    if limit is not None:
+        a_gelu = torch.where(
+            torch.isinf(a_gelu), a_gelu, a_gelu.clamp(min=None, max=limit)
+        )
+        a_linear = torch.where(
+            torch.isinf(a_linear), a_linear, a_linear.clamp(min=-limit, max=limit)
+        )
+    out_gelu = a_gelu * torch.sigmoid(1.702 * a_gelu)
+    return out_gelu * (a_linear + 1)
+
+
+class Phi3SmallMLP(nn.Module):
+    def __init__(self, prefix, config, weights):
+        super().__init__()
+
+        self.hidden_act = config.hidden_act
+        if self.hidden_act == "gegelu":
+            self.act = lambda x: gegelu(x, config.gegelu_limit)
+        elif "gelu" in self.hidden_act:
+            self.act = lambda x: torch.nn.functional.gelu(
+                x,
+                approximate=(
+                    "tanh"
+                    if self.hidden_act in ["gelu_fast", "gelu_pytorch_tanh"]
+                    else "none"
+                ),
+            )
+        else:
+            self.act = ACT2FN[self.hidden_act]
+
+        bias = getattr(config, "mlp_bias", False)
+        bias = True
+        self.up_proj = TensorParallelColumnLinear.load(
+            config, prefix=f"{prefix}.up_proj", weights=weights, bias=bias
+        )
+        self.down_proj = TensorParallelRowLinear.load(
+            config,
+            prefix=f"{prefix}.down_proj",
+            weights=weights,
+            bias=bias,
+        )
+        self.intermediate_size = (
+            config.intermediate_size // weights.process_group.size()
+        )
+
+        # TODO: This is a hotfix to be removed & properly refactored.
+        self.quantize = config.quantize
+
+    def forward(self, hidden_states):
+        if (
+            SYSTEM == "rocm"
+            and self.hidden_act == "silu"
+            and hidden_states.shape[0] == 1
+            and not self.quantize
+        ):
+            out = torch.empty(
+                hidden_states.shape[0],
+                self.intermediate_size,
+                dtype=hidden_states.dtype,
+                device="cuda",
+            )
+            _custom_C.LLMM_Silu(self.gate_up_proj.linear.weight, hidden_states, out, 8)
+            return self.down_proj(out)
+        else:
+            return self.down_proj(self.act(self.up_proj(hidden_states)))
+
+
+class FlashPhi3SmallLayer(nn.Module):
+    def __init__(self, prefix, config, weights):
+        super().__init__()
+        self.self_attn = FlashPhi3SmallAttention(
+            prefix=f"{prefix}.self_attn", config=config, weights=weights
+        )
+        self.mlp = Phi3SmallMLP(prefix=f"{prefix}.mlp", config=config, weights=weights)
+
+        self.input_layernorm = FastLayerNorm.load(
+            prefix=f"{prefix}.input_layernorm",
+            weights=weights,
+            eps=config.layer_norm_epsilon,
+        )
+        self.post_attention_layernorm = FastLayerNorm.load(
+            prefix=f"{prefix}.post_attention_layernorm",
+            weights=weights,
+            eps=config.layer_norm_epsilon,
+        )
+
+    def forward(
+        self,
+        hidden_states,
+        residual,
+        cos,
+        sin,
+        cu_seqlen_prefill,
+        kv_cache,
+        block_tables,
+        slots,
+        input_lengths,
+        max_s,
+    ):
+        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,
+        )
+
+        # faster post attention rms norm
+        normed_attn_res_output, attn_res = self.post_attention_layernorm(
+            attn_output, res
+        )
+
+        mlp_output = self.mlp(normed_attn_res_output)
+
+        return mlp_output, attn_res
+
+
+class FlashPhi3SmallModel(torch.nn.Module):
+    def __init__(self, prefix, config, weights):
+        super().__init__()
+
+        process_group = weights.process_group
+        self.tp_rank = process_group.rank()
+        self.tp_world_size = process_group.size()
+        self.layers = nn.ModuleList(
+            [
+                FlashPhi3SmallLayer(
+                    prefix=(
+                        f"model.layers.{layer_id}"
+                        if not prefix
+                        else f"{prefix}.model.layers.{layer_id}"
+                    ),
+                    config=config,
+                    weights=weights,
+                )
+                for layer_id in range(config.num_hidden_layers)
+            ]
+        )
+        self.norm = FastLayerNorm.load(
+            prefix=(
+                "model.final_layernorm"
+                if not prefix
+                else f"{prefix}.model.final_layernorm"
+            ),
+            weights=weights,
+            eps=config.layer_norm_epsilon,
+        )
+
+        self.gradient_checkpointing = False
+
+        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,
+        inputs_embeds: 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,
+        true_max_s: int,
+        prefill_cache_indices: Optional[torch.Tensor],
+    ) -> torch.Tensor:
+        hidden_states = inputs_embeds
+
+        # 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,
+            )
+
+        hidden_states, _ = self.norm(hidden_states, residual)
+
+        return hidden_states
+
+
+class FlashPhi3SmallForCausalLM(torch.nn.Module):
+    def __init__(self, prefix, config, weights):
+        super().__init__()
+
+        self.embed_tokens = TensorParallelEmbedding(
+            prefix=(
+                "model.embed_tokens" if not prefix else f"{prefix}.model.embed_tokens"
+            ),
+            weights=weights,
+        )
+        self.model = FlashPhi3SmallModel(prefix, config, weights)
+        if config.tie_word_embeddings:
+            suffix = "model.embed_tokens"
+        else:
+            suffix = "lm_head"
+
+        self.lm_head = SpeculativeHead.load(
+            config,
+            prefix=suffix if not prefix else f"{prefix}.{suffix}",
+            weights=weights,
+        )
+
+    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] = None,
+        lm_head_indices: Optional[torch.Tensor] = None,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        inputs_embeds = self.embed_tokens(input_ids)
+        hidden_states = self.model(
+            inputs_embeds,
+            position_ids,
+            cu_seqlen_prefill,
+            kv_cache,
+            block_tables,
+            slots,
+            input_lengths,
+            max_s,
+            true_max_s=max_s,
+            prefill_cache_indices=prefill_cache_indices,
+        )
+        if lm_head_indices is not None:
+            hidden_states = hidden_states[lm_head_indices]
+        logits, speculative_logits = self.lm_head(hidden_states)
+        return logits, speculative_logits

server/text_generation_server/models/flash_phi3small.py

@@ -0,0 +1,86 @@
+import torch
+import torch.distributed
+
+from opentelemetry import trace
+from transformers import AutoConfig, AutoTokenizer, GenerationConfig
+from typing import Optional
+
+from text_generation_server.models import FlashCausalLM
+from text_generation_server.models.custom_modeling.flash_phi3small_modeling import (
+    FlashPhi3SmallForCausalLM,
+)
+from text_generation_server.utils import (
+    initialize_torch_distributed,
+    weight_files,
+    Weights,
+)
+
+tracer = trace.get_tracer(__name__)
+
+from text_generation_server.utils.import_utils import SYSTEM
+
+
+class FlashPhi3Small(FlashCausalLM):
+    def __init__(
+        self,
+        model_id: str,
+        revision: Optional[str] = None,
+        quantize: Optional[str] = None,
+        speculator: Optional[str] = None,
+        dtype: Optional[torch.dtype] = None,
+        trust_remote_code: bool = False,
+    ):
+        self.process_group, rank, world_size = initialize_torch_distributed()
+        if torch.cuda.is_available():
+            device = torch.device(f"cuda:{rank}")
+            dtype = torch.float16 if dtype is None else dtype
+        elif SYSTEM == "xpu":
+            device = torch.device(f"xpu:{rank}")
+            dtype = torch.float16 if dtype is None else dtype
+        else:
+            raise NotImplementedError("FlashPhi3Small is only available on GPU")
+
+        tokenizer = AutoTokenizer.from_pretrained(
+            model_id,
+            revision=revision,
+            padding_side="left",
+            truncation_side="left",
+            trust_remote_code=trust_remote_code,
+        )
+        try:
+            generation_config = GenerationConfig.from_pretrained(
+                model_id, revision=revision, trust_remote_code=trust_remote_code
+            )
+            if isinstance(generation_config.eos_token_id, (list, set)):
+                # TODO Huge hack
+                tokenizer._eos_token_ids = set(generation_config.eos_token_id)
+        except Exception:
+            pass
+
+        config = AutoConfig.from_pretrained(
+            model_id, revision=revision, trust_remote_code=trust_remote_code
+        )
+        config.quantize = quantize
+        config.speculator = speculator
+
+        torch.distributed.barrier(group=self.process_group)
+
+        filenames = weight_files(model_id, revision=revision, extension=".safetensors")
+        weights = Weights(filenames, device, dtype, process_group=self.process_group)
+        if config.quantize in ["gptq", "awq", "exl2"]:
+            weights._set_gptq_params(model_id, revision)
+
+        prefix = ""
+        model = FlashPhi3SmallForCausalLM(prefix, config, weights)
+        torch.distributed.barrier(group=self.process_group)
+        super(FlashPhi3Small, self).__init__(
+            model=model,
+            tokenizer=tokenizer,
+            num_layers=len(model.model.layers),
+            num_kv_heads=model.model.num_key_value_heads,
+            head_size=model.model.head_size,
+            dtype=dtype,
+            device=device,
+            rank=rank,
+            world_size=world_size,
+        )