refactor internlm and internlm2 (#11274)

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

@@ -719,6 +719,10 @@ def _optimize_pre(model):
         # For stablelm-zephyr-3b and stablelm-2-zephyr-1_6b
         from ipex_llm.transformers.models.stablelm import merge_qkv
         model.apply(merge_qkv)
+    # for internlm
+    if model.config.model_type == "internlm":
+        from ipex_llm.transformers.models.internlm import merge_qkv
+        model.apply(merge_qkv)
     # for internlm-xcomposer2-vl
     if model.config.model_type == "internlmxcomposer2":
         from ipex_llm.transformers.models.internlm import pre_process_attn_and_mlp
@@ -1167,27 +1171,14 @@ def _optimize_post(model, lightweight_bmm=False):
         modeling_module_name = model.__class__.__module__
         module = importlib.import_module(modeling_module_name)
         from ipex_llm.transformers.models.internlm import internlm_attention_forward
+        convert_forward(model, module.InternLMAttention, internlm_attention_forward)
+        convert_forward(model, module.InternLMRMSNorm, llama_rms_norm_forward)
+    elif model.config.model_type == "internlm2":
+        modeling_module_name = model.__class__.__module__
+        module = importlib.import_module(modeling_module_name)
         from ipex_llm.transformers.models.internlm import internlm2_attention_forward
-        try:
+        convert_forward(model, module.InternLM2Attention, internlm2_attention_forward)
-            convert_forward(model,
+        convert_forward(model, module.InternLM2RMSNorm, llama_rms_norm_forward)
-                            module.InternLM2Attention,
-                            internlm2_attention_forward
-                            )
-        except:
-            convert_forward(model,
-                            module.InternLMAttention,
-                            internlm_attention_forward
-                            )
-        try:
-            convert_forward(model,
-                            module.InternLM2RMSNorm,
-                            llama_rms_norm_forward
-                            )
-        except:
-            convert_forward(model,
-                            module.InternLMRMSNorm,
-                            llama_rms_norm_forward
-                            )
     elif model.config.model_type == "internlmxcomposer2":
         modeling_module_name = model.model.__class__.__module__
         module = importlib.import_module(modeling_module_name)

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

@@ -42,20 +42,35 @@ from typing import Optional, Tuple, List
 import torch
 import torch.utils.checkpoint
 from torch import nn
-from ipex_llm.utils.common import invalidInputError
-from ipex_llm.transformers.models.utils import init_kv_cache, extend_kv_cache, \
-    append_kv_cache, is_enough_kv_cache_room_4_31
 from ipex_llm.transformers.models.utils import should_use_fuse_rope, apply_rotary_pos_emb
-from ipex_llm.transformers.models.utils import apply_rotary_pos_emb_no_cache_xpu
 from ipex_llm.transformers.models.utils import apply_rotary_pos_emb_cache_freq_xpu
 from ipex_llm.transformers.models.utils import use_quantize_kv_cache, restore_fp8_kv_cache
 from ipex_llm.transformers.models.utils import update_past_key_value
 from ipex_llm.transformers.models.utils import use_sdp, use_sdp_causal
 from einops import rearrange
-import os
 
 
-KV_CACHE_ALLOC_BLOCK_LENGTH = int(os.environ.get("KV_CACHE_ALLOC_BLOCK_LENGTH", 256))
+def merge_qkv(module: torch.nn.Module):
+    if module.__class__.__name__ == "InternLMAttention":
+        new_weight = torch.cat([
+            module.q_proj.weight.data,
+            module.k_proj.weight.data,
+            module.v_proj.weight.data,
+        ], dim=0)
+        new_bias = torch.cat([
+            module.q_proj.bias.data,
+            module.k_proj.bias.data,
+            module.v_proj.bias.data,
+        ], dim=-1)
+
+        qkv_proj = torch.nn.Linear(0, 0, bias=True)
+        qkv_proj.weight = torch.nn.Parameter(new_weight, requires_grad=False)
+        qkv_proj.bias = torch.nn.Parameter(new_bias, requires_grad=False)
+        qkv_proj.in_features = new_weight.size(1)
+        qkv_proj.out_features = new_weight.size(0)
+        module.qkv_proj = qkv_proj
+
+        del module.q_proj, module.k_proj, module.v_proj
 
 
 def internlm_attention_forward(
@@ -68,109 +83,69 @@ def internlm_attention_forward(
     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) \
+    qkv = self.qkv_proj(hidden_states)
-        .view(bsz, q_len, self.num_heads, self.head_dim) \
+    qkv = qkv.view(bsz, q_len, self.num_heads * 3, self.head_dim)
-        .transpose(1, 2)
+    qkv = qkv.transpose(1, 2)
-    key_states = self.k_proj(hidden_states) \
+    query_states, key_states, value_states = qkv.split([self.num_heads,
-        .view(bsz, q_len, self.num_heads, self.head_dim) \
+                                                        self.num_heads,
-        .transpose(1, 2)
+                                                        self.num_heads], dim=1)
-    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]
-    enough_kv_room = True
     if past_key_value is not None:
-        enough_kv_room = is_enough_kv_cache_room_4_31(past_key_value, seq_len=kv_seq_len)
         kv_seq_len += past_key_value[0].shape[-2]
+
+    # IPEX-LLM OPT: fuse rope
     if should_use_fuse_rope(hidden_states, position_ids, self.training):
-        query_states, key_states = apply_rotary_pos_emb_no_cache_xpu(query_states,
+        import xe_addons
-                                                                     key_states,
+        xe_addons.rotary_half_inplaced(self.rotary_emb.inv_freq, position_ids,
-                                                                     position_ids,
+                                       query_states, key_states)
-                                                                     "internlm")
     else:
         cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
         query_states, key_states = apply_rotary_pos_emb(
-            query_states,
+            query_states, key_states, cos, sin, position_ids, "internlm"
-            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 not enough_kv_room:
-            # 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
 
+    # IPEX-LLM OPT: kv cache and quantzie kv cache
+    use_quantize_kv = use_quantize_kv_cache(self.qkv_proj, hidden_states)
+    key_states, value_states = update_past_key_value(
+        past_key_value, key_states, value_states,
+        kv_seq_len, use_quantize_kv, hidden_states.device
+    )
     past_key_value = (key_states, value_states) if use_cache else None
 
-    attn_weights = torch.matmul(query_states,
+    # IPEX-LLM OPT: sdp
-                                key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+    attn_weights = None
+    if use_sdp(q_len, kv_seq_len, self.head_dim, query_states):
+        import xe_addons
+        if use_quantize_kv:
+            attn_output = xe_addons.sdp_fp8(query_states, key_states, value_states,
+                                            attention_mask)
+        else:
+            attn_output = xe_addons.sdp(query_states, key_states, value_states, attention_mask)
+    elif use_sdp_causal(q_len, kv_seq_len, self.head_dim, query_states, self.training):
+        import xe_addons
+        if use_quantize_kv:
+            attn_output = xe_addons.sdp_fp8_causal(query_states, key_states,
+                                                   value_states, attention_mask)
+        else:
+            attn_output = xe_addons.sdp_causal(query_states, key_states,
+                                               value_states, attention_mask)
+    else:
+        if use_quantize_kv:
+            key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
+                                                            query_states.dtype)
 
-    if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+        attn_weights = torch.matmul(query_states,
-        invalidInputError(
+                                    key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
-            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 is not None:
-        if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+            attn_weights = attn_weights + attention_mask
-            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
+        # upcast attention to fp32
-    attn_weights = nn.functional.softmax(attn_weights,
+        attn_weights = nn.functional.softmax(attn_weights,
-                                         dim=-1, dtype=torch.float32).to(query_states.dtype)
+                                             dim=-1, dtype=torch.float32).to(query_states.dtype)
-    attn_output = torch.matmul(attn_weights, value_states)
+        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)
@@ -229,62 +204,60 @@ def internlm2_attention_forward(
     kv_seq_len = key_states.shape[-2]
     if past_key_value is not None:
         kv_seq_len += past_key_value[0].shape[-2]
+
+    # IPEX-LLM OPT: fuse rope
     if should_use_fuse_rope(hidden_states, position_ids, self.training):
-        query_states, key_states = apply_rotary_pos_emb_no_cache_xpu(query_states,
+        import xe_addons
-                                                                     key_states,
+        xe_addons.rotary_half_inplaced(self.rotary_emb.inv_freq, position_ids,
-                                                                     position_ids,
+                                       query_states, key_states)
-                                                                     "internlm")
     else:
         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)
         query_states, key_states = apply_rotary_pos_emb(
-            query_states,
+            query_states, key_states, cos, sin, position_ids, "internlm"
-            key_states,
+        )
-            cos,
-            sin,
-            position_ids,
-            "internlm")
-
-    if past_key_value is not None:
-        # reuse k, v, self_attention
-        key_states = torch.cat([past_key_value[0], key_states], dim=2)
-        value_states = torch.cat([past_key_value[1], value_states], dim=2)
 
+    # IPEX-LLM OPT: kv cache and quantzie kv cache
+    use_quantize_kv = use_quantize_kv_cache(self.wqkv, hidden_states)
+    key_states, value_states = update_past_key_value(
+        past_key_value, key_states, value_states,
+        kv_seq_len, use_quantize_kv, hidden_states.device
+    )
     past_key_value = (key_states, value_states) if use_cache else None
 
-    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    # IPEX-LLM OPT: sdp
-    value_states = repeat_kv(value_states, self.num_key_value_groups)
+    attn_weights = None
+    if use_sdp(q_len, kv_seq_len, self.head_dim, query_states):
+        import xe_addons
+        if use_quantize_kv:
+            attn_output = xe_addons.sdp_fp8(query_states, key_states, value_states,
+                                            attention_mask)
+        else:
+            attn_output = xe_addons.sdp(query_states, key_states, value_states, attention_mask)
+    elif use_sdp_causal(q_len, kv_seq_len, self.head_dim, query_states, self.training):
+        import xe_addons
+        if use_quantize_kv:
+            attn_output = xe_addons.sdp_fp8_causal(query_states, key_states,
+                                                   value_states, attention_mask)
+        else:
+            attn_output = xe_addons.sdp_causal(query_states, key_states,
+                                               value_states, attention_mask)
+    else:
+        if use_quantize_kv:
+            key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
+                                                            query_states.dtype)
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
 
-    attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+        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):
+        if attention_mask is not None:
-        invalidInputError(
+            attn_weights = attn_weights + attention_mask
-            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:
+        # upcast attention to fp32
-        if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+        attn_weights = nn.functional.softmax(attn_weights,
-            invalidInputError(
+                                             dim=-1, dtype=torch.float32).to(query_states.dtype)
-                False,
+        attn_output = torch.matmul(attn_weights, value_states)
-                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
-
-    # 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).contiguous()
     attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)