From f1410d68233619bf6981e4d122f39b67c137bee1 Mon Sep 17 00:00:00 2001 From: Xin Qiu Date: Thu, 13 Jun 2024 18:06:04 +0800 Subject: [PATCH] refactor chatglm4 (#11301) * glm4 * remove useless code * stype * add rope_ratio * update * fix fp16 * fix style --- .../ipex_llm/transformers/models/chatglm4.py | 283 ++++++++---------- 1 file changed, 128 insertions(+), 155 deletions(-) diff --git a/python/llm/src/ipex_llm/transformers/models/chatglm4.py b/python/llm/src/ipex_llm/transformers/models/chatglm4.py index 00a69dcd821..86aeaba134f 100644 --- a/python/llm/src/ipex_llm/transformers/models/chatglm4.py +++ b/python/llm/src/ipex_llm/transformers/models/chatglm4.py @@ -20,13 +20,12 @@ import torch from typing import Optional, Tuple, Union, List, Callable, Dict, Any import torch.nn.functional as F -from ipex_llm.transformers.models.utils import init_kv_cache, extend_kv_cache, append_kv_cache -from ipex_llm.transformers.models.utils import use_quantize_kv_cache, apply_ipex_rotate_every_two -from ipex_llm.transformers.models.utils import use_sdp -from ipex_llm.transformers.models.chatglm2 import should_split_qkv_tensor, glm_sdpa -from ipex_llm.transformers.models.chatglm2 import split_tensor_along_last_dim +from ipex_llm.transformers.models.utils import restore_fp8_kv_cache, update_past_key_value +from ipex_llm.transformers.models.utils import use_quantize_kv_cache, use_sdp, use_sdp_causal +from ipex_llm.transformers.models.utils import should_use_fuse_rope, apply_rotary_pos_emb +from ipex_llm.transformers.models.chatglm2 import repeat_kv from transformers.modeling_outputs import BaseModelOutputWithPast - +import math import os @@ -97,6 +96,18 @@ def chatglm4_model_forward_internal( past_key_values, padding_mask=attention_mask) + # ipex-llm changes begin + # 1. replace `rotary_pos_emb` with `inv_freq` and `position_ids` + # 2. generate `causal_mask` and replace `full_attention_mask` with it + if position_ids is None: + if past_key_values is None: + position_ids = torch.arange(seq_length, dtype=torch.int64, device=inputs_embeds.device) + else: + kv_length = past_key_values[0][0].size(2) + position_ids = torch.arange(kv_length, kv_length + seq_length, + dtype=torch.int64, device=inputs_embeds.device) + position_ids = position_ids.repeat(batch_size, 1) + use_fuse_rope = input_ids.device.type == "xpu" use_fuse_rope = use_fuse_rope and not self.training @@ -117,11 +128,31 @@ def chatglm4_model_forward_internal( sin = torch.repeat_interleave(sin[:, :, None, :], 2, 3) rotary_pos_emb = (cos, sin) - # Run encoder. + # `full_attention_mask` is not None only when + # `past_key_values` is not None and `seq_length` > 1 + if full_attention_mask is not None: + causal_mask = torch.zeros([batch_size, 1, seq_length, full_attention_mask.size(-1)], + dtype=inputs_embeds.dtype, device=inputs_embeds.device) + mask_value = torch.finfo(inputs_embeds.dtype).min + causal_mask.masked_fill_(full_attention_mask, mask_value) + elif self.training or (inputs_embeds.device.type != "xpu" and past_key_values is None): + full_attention_mask = self.get_masks(input_ids, + past_key_values, + padding_mask=attention_mask) + causal_mask = torch.zeros([batch_size, 1, seq_length, full_attention_mask.size(-1)], + dtype=inputs_embeds.dtype, device=inputs_embeds.device) + mask_value = torch.finfo(inputs_embeds.dtype).min + causal_mask.masked_fill_(full_attention_mask, mask_value) + else: + causal_mask = None + hidden_states, presents, all_hidden_states, all_self_attentions = self.encoder( - inputs_embeds, full_attention_mask, rotary_pos_emb=rotary_pos_emb, + inputs_embeds, causal_mask, + rotary_pos_emb=rotary_pos_emb, kv_caches=past_key_values, use_cache=use_cache, output_hidden_states=output_hidden_states ) + # ipex-llm changes end + if presents is not None and type(presents) is torch.Tensor: presents = presents.split(1, dim=0) presents = list(presents) @@ -141,7 +172,6 @@ def chatglm4_model_forward_internal( ) -@torch.jit.script def apply_rotary_pos_emb(x: torch.Tensor, rope_cache: torch.Tensor) -> torch.Tensor: # x: [b, np, sq, hn] b, np, sq, hn = x.size(0), x.size(1), x.size(2), x.size(3) @@ -165,167 +195,110 @@ def apply_rotary_pos_emb(x: torch.Tensor, rope_cache: torch.Tensor) -> torch.Ten def chatglm4_attention_forward( self, hidden_states, attention_mask, rotary_pos_emb, kv_cache=None, use_cache=True ): - # hidden_states: [sq, b, h] - - # ================================================= - # Pre-allocate memory for key-values for inference. - # ================================================= - # ===================== - # Query, Key, and Value - # ===================== - - # Attention heads [sq, b, h] --> [sq, b, (np * 3 * hn)] - device = hidden_states.device - mixed_x_layer = self.query_key_value(hidden_states) - - if self.multi_query_attention: - (query_layer, key_layer, value_layer) = mixed_x_layer.split( - [ - self.num_attention_heads_per_partition * self.hidden_size_per_attention_head, - self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head, - self.num_multi_query_groups_per_partition * self.hidden_size_per_attention_head, - ], - dim=-1, - ) - query_layer = query_layer.view( - query_layer.size()[:-1] + (self.num_attention_heads_per_partition, - self.hidden_size_per_attention_head) - ) - key_layer = key_layer.view( - key_layer.size()[:-1] + (self.num_multi_query_groups_per_partition, - self.hidden_size_per_attention_head) - ) - value_layer = value_layer.view( - value_layer.size()[:-1] - + (self.num_multi_query_groups_per_partition, self.hidden_size_per_attention_head) - ) - else: - new_tensor_shape = mixed_x_layer.size()[:-1] + (self.num_attention_heads_per_partition, - 3 * self.hidden_size_per_attention_head) - mixed_x_layer = mixed_x_layer.view(*new_tensor_shape) + # hidden_states: [b, sq, h] + bsz, q_len, _ = hidden_states.size() - # [sq, b, np, 3 * hn] --> 3 [sq, b, np, hn] - (query_layer, key_layer, value_layer) = split_tensor_along_last_dim(mixed_x_layer, 3) + # past_key_value: [bsz, n_kv_head, seq_len, head_dim] + past_key_value = None if kv_cache is None else (kv_cache[0], + kv_cache[1]) - # [b, sq, np, hn] -> [b, np, sq, hn] - query_layer, key_layer, value_layer = [k.transpose(1, 2) - for k in [query_layer, key_layer, value_layer]] + n_head = self.num_attention_heads_per_partition + n_kv_head = self.num_multi_query_groups_per_partition if self.multi_query_attention else n_head + head_dim = self.hidden_size_per_attention_head + + qkv = self.query_key_value(hidden_states) + # [bs, q_len, np * 3 * hn] -> [bsz, n_head, seq_len, head_dim] + qkv = qkv.view(bsz, q_len, n_head + 2 * n_kv_head, head_dim) + + query_states, key_states, value_states = qkv.split([n_head, + n_kv_head, + n_kv_head], dim=2) + + kv_seq_len = key_states.shape[1] + if past_key_value is not None: + kv_seq_len += past_key_value[0].shape[2] - # apply relative positional encoding (rotary embedding) if isinstance(rotary_pos_emb, tuple) and len(rotary_pos_emb) == 2: # use_fuse_rope, see chatglm4_model_forward cos, sin = rotary_pos_emb rot_dim = cos.shape[-1] - query_layer = query_layer.transpose(1, 2) - key_layer = key_layer.transpose(1, 2) - query_layer_cur = query_layer[..., :rot_dim] - key_layer_cur = key_layer[..., :rot_dim] + query_layer_cur = query_states[..., :rot_dim] + key_layer_cur = key_states[..., :rot_dim] # ipex_llm's apply_rotary_embedding can change the origin storage, # so query_layer will get the result directly. torch.ops.torch_ipex.apply_rotary_embedding(query_layer_cur, sin, cos, query_layer_cur) torch.ops.torch_ipex.apply_rotary_embedding(key_layer_cur, sin, cos, key_layer_cur) - query_layer = query_layer.transpose(1, 2) - key_layer = key_layer.transpose(1, 2) + query_states = query_states.transpose(1, 2) + key_states = key_states.transpose(1, 2) + value_states = value_states.transpose(1, 2) elif rotary_pos_emb is not None: - query_layer = apply_rotary_pos_emb(query_layer, rotary_pos_emb) - key_layer = apply_rotary_pos_emb(key_layer, rotary_pos_emb) - - cur_length, batch_size = query_layer.shape[2], query_layer.shape[0] - - # adjust key and value for inference - if kv_cache is not None and use_cache: - cache_k, cache_v = kv_cache - past_length = cache_k.size(2) - - if cache_k.stride()[1] < (past_length + cur_length) * cache_k.size(3): - max_cache_length = past_length + cur_length + KV_CACHE_ALLOC_BLOCK_LENGTH - new_cache_k, new_cache_v = extend_kv_cache(batch_size, - key_layer.size(1), - self.hidden_size_per_attention_head, - past_length, - max_cache_length, - dtype=query_layer.dtype, - device=device) - new_cache_k[:] = cache_k - new_cache_v[:] = cache_v - cache_k = new_cache_k - cache_v = new_cache_v - - key_layer, value_layer = append_kv_cache(cache_k, cache_v, key_layer, value_layer) + query_states = query_states.transpose(1, 2) + key_states = key_states.transpose(1, 2) + value_states = value_states.transpose(1, 2) + query_states = apply_rotary_pos_emb(query_states, rotary_pos_emb) + key_states = apply_rotary_pos_emb(key_states, rotary_pos_emb) + + # IPEX-LLM OPT: kv cache and quantize kv + use_quantize_kv = use_quantize_kv_cache(self.query_key_value, 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 + ) if use_cache: - if kv_cache is None: - kv_cache = torch.cat((key_layer.unsqueeze(0).unsqueeze(0), - value_layer.unsqueeze(0).unsqueeze(0)), dim=1) + if past_key_value is None: + past_key_value = torch.cat((key_states.unsqueeze(0).unsqueeze(0), + value_states.unsqueeze(0).unsqueeze(0)), dim=1) else: - kv_cache = (key_layer, value_layer) + past_key_value = (key_states, value_states) else: - kv_cache = None - - if self.multi_query_attention: - key_layer = key_layer.unsqueeze(2) - key_layer = key_layer.expand( - -1, -1, - self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, - -1, -1 - ) - key_layer = key_layer.contiguous().view( - key_layer.size()[:1] + (self.num_attention_heads_per_partition,) + key_layer.size()[3:] - ) - value_layer = value_layer.unsqueeze(2) - value_layer = value_layer.expand( - -1, -1, - self.num_attention_heads_per_partition // self.num_multi_query_groups_per_partition, - -1, -1 - ) - value_layer = value_layer.contiguous().view( - value_layer.size()[:1] + - (self.num_attention_heads_per_partition,) + value_layer.size()[3:] - ) - - # ================================== - # core attention computation - # ================================== - - context_layer = core_attn_forward(query_layer, key_layer, value_layer, attention_mask) - - # ================= - # Output. [sq, b, h] - # ================= - - output = self.dense(context_layer) - - return output, kv_cache - - -def core_attn_forward(query_layer, key_layer, value_layer, attention_mask): - L, S = query_layer.shape[2], key_layer.shape[2] - batch_size, n_head, seq_len, head_dim = query_layer.shape - if attention_mask is None and L == S: - if should_split_qkv_tensor(query_layer, batch_size, n_head, seq_len): - # split second dim to block size = 8 - block_size = 8 - query_layer = query_layer.to(key_layer.dtype) - query_split = torch.split(query_layer, block_size, dim=1) - key_split = torch.split(key_layer, block_size, dim=1) - value_split = torch.split(value_layer, block_size, dim=1) - results = [] - for q, k, v in zip(query_split, key_split, value_split): - result = glm_sdpa(q, k, v, is_causal=True) - results.append(result) - context_layer = torch.cat(results, dim=1) + past_key_value = None + + # IPEX-LLM OPT: sdp + attn_weights = None + if use_sdp(q_len, kv_seq_len, 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, 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) + elif query_states.device.type == "cpu": + # repeat k/v heads if n_kv_heads < n_heads + key_states = repeat_kv(key_states, n_head // n_kv_head) + value_states = repeat_kv(value_states, n_head // n_kv_head) + if q_len == kv_seq_len: + attn_output = torch.nn.functional.scaled_dot_product_attention( + query_states, key_states, value_states, is_causal=True + ) else: - context_layer = glm_sdpa(query_layer, - key_layer, - value_layer, - is_causal=True) + attn_output = torch.nn.functional.scaled_dot_product_attention( + query_states, key_states, value_states, attention_mask + ) else: - context_layer = glm_sdpa(query_layer, - key_layer, - value_layer, - attention_mask) - context_layer = context_layer.transpose(1, 2).contiguous() - new_context_layer_shape = context_layer.size()[:-2] + (-1,) - context_layer = context_layer.reshape(*new_context_layer_shape) - - return context_layer + if use_quantize_kv: + key_states, value_states = restore_fp8_kv_cache(key_states, value_states, + query_states.dtype) + # repeat k/v heads if n_kv_heads < n_heads + key_states = repeat_kv(key_states, n_head // n_kv_head) + value_states = repeat_kv(value_states, n_head // n_kv_head) + attn_weights = torch.matmul(query_states / math.sqrt(head_dim), + key_states.transpose(2, 3)).to(value_states.dtype) + if attention_mask is not None: + attn_weights = attn_weights + attention_mask + attn_weights = torch.nn.functional.softmax(attn_weights, dim=-1, + dtype=torch.float32).to(value_states.dtype) + attn_output = torch.matmul(attn_weights, value_states) + + # context_layer's shape: [bsz, n_head, seq_len, head_dim] -> [seq_len, bsz, n_head * head_dim] + attn_output = attn_output.transpose(1, 2).contiguous().view(bsz, q_len, n_head * head_dim) + output = self.dense(attn_output) + + return output, past_key_value