[WIP] LLM: add kv cache support for internlm. (#9036)

* LLM: add kv cache support for internlm

* add internlm apply_rotary_pos_emb

* fix.

* fix style.

python/llm/src/bigdl/llm/transformers/convert.py

@@ -272,4 +272,12 @@ def optimize(model):
                         transformers.models.gpt_neox.modeling_gpt_neox.GPTNeoXAttention,
                         gptneox_attention_forward
                         )
+    elif model.config.model_type == "internlm":
+        modeling_module_name = model.__class__.__module__
+        module = importlib.import_module(modeling_module_name)
+        from bigdl.llm.transformers.models.internlm import internlm_attention_forward
+        convert_forward(model,
+                        module.InternLMAttention,
+                        internlm_attention_forward
+                        )
     return model

python/llm/src/bigdl/llm/transformers/models/internlm.py

@@ -0,0 +1,168 @@
+#
+# Copyright 2016 The BigDL Authors.
+#
+# 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.
+#
+# Some parts of this file is adapted from
+# https://huggingface.co/internlm/internlm-chat-7b/blob/659ed911eec1e26810f9854f19c5ec27854e9cf3/modeling_internlm.py
+# which is licensed under Apache License 2.0:
+#
+# 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.
+""" PyTorch InternLM model."""
+import math
+from typing import Optional, Tuple
+
+import torch
+import torch.utils.checkpoint
+from torch import nn
+from bigdl.llm.utils.common import invalidInputError
+from bigdl.llm.transformers.models.utils import init_kv_cache, extend_kv_cache, append_kv_cache
+from bigdl.llm.transformers.models.utils import apply_rotary_pos_emb
+
+
+KV_CACHE_ALLOC_BLOCK_LENGTH = 256
+
+
+def internlm_attention_forward(
+    self,
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor]=None,
+    position_ids: Optional[torch.LongTensor]=None,
+    past_key_value: Optional[Tuple[torch.Tensor]]=None,
+    output_attentions: bool=False,
+    use_cache: bool=False,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    bsz, q_len, _ = hidden_states.size()
+    device = hidden_states.device
+    query_states = self.q_proj(hidden_states) \
+        .view(bsz, q_len, self.num_heads, self.head_dim) \
+        .transpose(1, 2)
+    key_states = self.k_proj(hidden_states) \
+        .view(bsz, q_len, self.num_heads, self.head_dim) \
+        .transpose(1, 2)
+    value_states = self.v_proj(hidden_states) \
+        .view(bsz, q_len, self.num_heads, self.head_dim) \
+        .transpose(1, 2)
+
+    kv_seq_len = key_states.shape[-2]
+    if past_key_value is not None:
+        kv_seq_len += past_key_value[0].shape[-2]
+    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+    query_states, key_states = apply_rotary_pos_emb(
+        query_states,
+        key_states,
+        cos,
+        sin,
+        position_ids,
+        "internlm"
+    )
+    # [bsz, nh, t, hd]
+
+    if past_key_value is not None:
+        # reuse k, v, self_attention
+        cache_k = past_key_value[0]
+        cache_v = past_key_value[1]
+        if cache_k.stride()[1] <= cache_k.size(2) * cache_k.size(3):
+            # allocate new
+            new_cache_k, new_cache_v = extend_kv_cache(
+                bsz,
+                self.num_heads,
+                self.head_dim,
+                cache_k.size(2),
+                kv_seq_len + KV_CACHE_ALLOC_BLOCK_LENGTH,
+                dtype=cache_k.dtype,
+                device=device
+            )
+            new_cache_k[:] = cache_k
+            new_cache_v[:] = cache_v
+            cache_k = new_cache_k
+            cache_v = new_cache_v
+
+        key_states, value_states = append_kv_cache(cache_k, cache_v, key_states, value_states)
+
+    elif use_cache:
+        max_cache_length = kv_seq_len + KV_CACHE_ALLOC_BLOCK_LENGTH
+        new_key_states, new_value_states = init_kv_cache(
+            bsz,
+            self.num_heads,
+            self.head_dim,
+            kv_seq_len,
+            max_cache_length,
+            dtype=key_states.dtype,
+            device=device
+        )
+        new_key_states[:] = key_states
+        new_value_states[:] = value_states
+        key_states = new_key_states
+        value_states = new_value_states
+
+    past_key_value = (key_states, value_states) if use_cache else None
+
+    attn_weights = torch.matmul(query_states,
+                                key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+    if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+        invalidInputError(
+            False,
+            f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, "
+            f"but is {attn_weights.size()}"
+        )
+
+    if attention_mask is not None:
+        if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+            invalidInputError(
+                False,
+                f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, "
+                f"but is {attention_mask.size()}"
+            )
+        attn_weights = attn_weights + attention_mask
+        attn_weights = torch.max(attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min))
+
+    # upcast attention to fp32
+    attn_weights = nn.functional.softmax(attn_weights,
+                                         dim=-1, dtype=torch.float32).to(query_states.dtype)
+    attn_output = torch.matmul(attn_weights, value_states)
+
+    if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+        invalidInputError(
+            False,
+            f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, "
+            f"but is {attn_output.size()}"
+        )
+
+    attn_output = attn_output.transpose(1, 2)
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+    attn_output = self.o_proj(attn_output)
+
+    if not output_attentions:
+        attn_weights = None
+
+    return attn_output, attn_weights, past_key_value

python/llm/src/bigdl/llm/transformers/models/utils.py

@@ -71,7 +71,7 @@ def rotate_every_two(x):
 
 
 def apply_rotary_pos_emb(q, k, cos, sin, position_ids, model_family):
-    if model_family in ["llama", "baichuan"]:
+    if model_family in ["llama", "baichuan", "internlm"]:
         # The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
         cos = cos.squeeze(1).squeeze(0)  # [seq_len, dim]
         sin = sin.squeeze(1).squeeze(0)  # [seq_len, dim]