Merge branch 'main' into sync/npu

colossalai/shardformer/modeling/llama.py

@@ -2,6 +2,7 @@ import warnings
 from typing import List, Optional, Tuple
 
 import torch
+import torch.nn.functional as F
 from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 from transformers.modeling_outputs import (
     BaseModelOutputWithPast,
@@ -12,6 +13,9 @@ from transformers.models.llama.modeling_llama import LlamaForCausalLM, LlamaForS
 from transformers.utils import logging
 
 from colossalai.pipeline.stage_manager import PipelineStageManager
+from colossalai.shardformer.shard import ShardConfig
+
+from ..layer import cross_entropy_1d
 
 try:
     from transformers.models.llama.modeling_llama import _prepare_4d_causal_attention_mask
@@ -42,6 +46,7 @@ class LlamaPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
     ):
         logger = logging.get_logger(__name__)
 
@@ -200,6 +205,7 @@ class LlamaPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
     ):
         r"""
         Args:
@@ -269,11 +275,18 @@ class LlamaPipelineForwards:
                 shift_labels = labels[..., 1:].contiguous()
                 # Flatten the tokens
                 loss_fct = CrossEntropyLoss()
-                shift_logits = shift_logits.view(-1, self.config.vocab_size)
                 shift_labels = shift_labels.view(-1)
                 # Enable model parallelism
                 shift_labels = shift_labels.to(shift_logits.device)
-                loss = loss_fct(shift_logits, shift_labels)
+                if shard_config.enable_tensor_parallelism:
+                    new_vocab_size = logits.shape[-1]
+                    shift_logits = shift_logits.view(-1, new_vocab_size)
+                    loss = cross_entropy_1d(
+                        shift_logits, shift_labels, process_group=shard_config.tensor_parallel_process_group
+                    )
+                else:
+                    shift_logits = shift_logits.view(-1, self.config.vocab_size)
+                    loss = loss_fct(shift_logits, shift_labels)
 
             if not return_dict:
                 output = (logits,) + outputs[1:]
@@ -306,6 +319,7 @@ class LlamaPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
     ):
         r"""
         labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
@@ -403,7 +417,7 @@ class LlamaPipelineForwards:
             return {"hidden_states": hidden_states}
 
 
-def get_llama_flash_attention_forward():
+def get_llama_flash_attention_forward(shard_config: ShardConfig):
     from transformers.models.llama.modeling_llama import LlamaAttention, apply_rotary_pos_emb
 
     from colossalai.kernel import AttnMaskType, ColoAttention
@@ -459,14 +473,13 @@ def get_llama_flash_attention_forward():
 
         flash_attention_mask = None
         attn_mask_type = AttnMaskType.causal
-        if attention_mask != None:
+        if not getattr(shard_config, "causal_lm", False) and attention_mask != None:
             if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
                 raise ValueError(
                     f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
                 )
             flash_attention_mask = ~(attention_mask[:, :, -1].squeeze(1).to(torch.bool)).contiguous()
-            if not torch.all(flash_attention_mask):
-                attn_mask_type = AttnMaskType.paddedcausal
+            attn_mask_type = AttnMaskType.paddedcausal
 
         attention = ColoAttention(embed_dim=self.hidden_size, num_heads=self.num_heads)
         attn_output = attention(
@@ -483,3 +496,108 @@ def get_llama_flash_attention_forward():
         return attn_output, None, past_key_value
 
     return forward
+
+
+def get_lm_forward_with_dist_cross_entropy(shard_config: ShardConfig):
+    from transformers import LlamaForCausalLM
+
+    def forward(
+        self: LlamaForCausalLM,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, LlamaForCausalLM
+
+        >>> model = LlamaForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
+        >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
+
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        if self.config.pretraining_tp > 1:
+            lm_head_slices = self.lm_head.weight.split(self.vocab_size // self.config.pretraining_tp, dim=0)
+            logits = [F.linear(hidden_states, lm_head_slices[i]) for i in range(self.config.pretraining_tp)]
+            logits = torch.cat(logits, dim=-1)
+        else:
+            logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            if shard_config.enable_tensor_parallelism:
+                new_vocab_size = logits.shape[-1]
+                shift_logits = shift_logits.view(-1, new_vocab_size)
+                loss = cross_entropy_1d(
+                    shift_logits, shift_labels, process_group=shard_config.tensor_parallel_process_group
+                )
+            else:
+                shift_logits = shift_logits.view(-1, self.config.vocab_size)
+                loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    return forward