Support multi-gpu inference

configs/accelerator_multigpu_config.yaml

@@ -8,11 +8,11 @@ machine_rank: 0
 main_training_function: main
 mixed_precision: bf16
 num_machines: 1
-num_processes: 2
+num_processes: 1
 rdzv_backend: static
 same_network: true
 tpu_env: []
 tpu_use_cluster: false
 tpu_use_sudo: false
 use_cpu: false
-main_process_port: 11000
+main_process_port: 11002

configs/ds_config_zero3.json

@@ -15,21 +15,16 @@
     "zero_optimization": {
         "stage": 3,
         "offload_optimizer": {
-            "device": "cpu",
-            "pin_memory": true
-        },
-        "offload_param": {
-            "device": "cpu",
-            "pin_memory": true
+            "device": "cpu"
         },
         "overlap_comm": true,
         "contiguous_gradients": true,
         "sub_group_size": 1e9,
         "reduce_bucket_size": "auto",
         "stage3_prefetch_bucket_size": "auto",
         "stage3_param_persistence_threshold": "auto",
-        "stage3_max_live_parameters": 1e9,
-        "stage3_max_reuse_distance": 1e9,
+        "stage3_max_live_parameters": 2e10,
+        "stage3_max_reuse_distance": 2e10,
         "stage3_gather_16bit_weights_on_model_save": true
     },
 

examples/chatbot.py

@@ -61,6 +61,7 @@ def main():
         tune_strategy='none',
         ds_config=ds_config,
         device=pipeline_args.device,
+        use_accelerator=True,
     )
 
     # We don't need input data, we will read interactively from stdin
@@ -120,7 +121,7 @@ def main():
         print("Bot: ", end="")
         print_index = 0
 
-        token_per_step = 4
+        token_per_step = 100
 
         for response, flag_break in inferencer.stream_inference(
             context=context,

scripts/run_chatbot.sh

@@ -9,8 +9,13 @@ if [ $# -ge 2 ]; then
   lora_args="--lora_model_path $2"
 fi
 
-CUDA_VISIBLE_DEVICES=0 \
-  deepspeed examples/chatbot.py \
-      --deepspeed configs/ds_config_chatbot.json \
-      --model_name_or_path ${model} \
-      ${lora_args}
+    # --temperature 0.7 \
+accelerate launch --config_file configs/accelerator_multigpu_config.yaml \
+  examples/chatbot.py \
+    --deepspeed configs/ds_config_chatbot.json \
+    --model_name_or_path ${model} \
+    --use_accelerator True \
+    --max_new_tokens 256 \
+    --temperature 1.0 \
+    --end_string "#" \
+    ${lora_args}

src/lmflow/args.py

@@ -798,6 +798,10 @@ class InferencerArguments:
             "help": "whether turn on true random sampling during inference."
         },
     )
+    use_accelerator: bool = field(
+        default=False, metadata={"help": "Whether to use Huggingface Accelerator instead of Deepspeed"},
+    )
+
 
 @dataclass
 class RaftAlignerArguments(TrainingArguments):

src/lmflow/models/hf_decoder_model.py

@@ -290,7 +290,6 @@ def __init__(
                         revision=model_args.model_revision,
                         use_auth_token=True if model_args.use_auth_token else None,
                         torch_dtype=torch_dtype,
-                        device_map=device_map,
                         trust_remote_code = model_args.trust_remote_code,
                     )
                 #for deepspeed zero3, we don't need to specify device_map

src/lmflow/pipeline/inferencer.py

@@ -16,6 +16,7 @@
 from typing import Dict, List
 from concurrent.futures import ThreadPoolExecutor
 
+from accelerate import Accelerator
 from transformers import AutoConfig
 import torch.distributed as dist
 import torch.nn.functional as F
@@ -81,6 +82,10 @@ def __init__(self, model_args, data_args, inferencer_args):
             print("Error in setting hidden size, use the default size 1024")
             self.model_hidden_size = 1024 # gpt2 seems do not have hidden_size in config
 
+        if inferencer_args.use_accelerator:
+            self.accelerator = Accelerator()
+            self.accelerator.wait_for_everyone()
+
 
     def create_dataloader(self, dataset: Dataset):
         r"""Batchlize dataset and format it to dataloader.
@@ -161,7 +166,7 @@ def inference(
                 input = current_batch['input']
                 input['text'] = prompt_structure.format(input=input['text'])
 
-            if 'images' in input and isinstance(input['images'], list):
+            if False and 'images' in input and isinstance(input['images'], list):
                 input['images'] = np.array(input['images'])
             if remove_image_flag:
                 # remove the image flag <ImageHere> in tokenization;
@@ -219,17 +224,33 @@ def inference(
                     raise NotImplementedError(
                         f"device \"{self.inferencer_args.device}\" is not supported"
                     )
+
+                if self.inferencer_args.use_accelerator:
+                    inputs = inputs.to(self.accelerator.device)
+
+
             if remove_image_flag:
                 inputs["image_token_indexes"] = image_token_indexes
                 inputs["one_sample_multiple_images"] = True
 
-            outputs = model.inference(
-                inputs,
-                max_new_tokens=max_new_tokens,
-                temperature=self.inferencer_args.temperature,
-                repetition_penalty=self.inferencer_args.repetition_penalty,
-                do_sample=self.inferencer_args.do_sample,
-            )
+            if self.inferencer_args.use_accelerator:
+                with self.accelerator.autocast():
+                    outputs = model.inference(
+                        inputs,
+                        max_new_tokens=max_new_tokens,
+                        temperature=self.inferencer_args.temperature,
+                        repetition_penalty=self.inferencer_args.repetition_penalty,
+                        do_sample=self.inferencer_args.do_sample,
+                        use_accelerator=True,
+                    )
+            else:
+                outputs = model.inference(
+                    inputs,
+                    max_new_tokens=max_new_tokens,
+                    temperature=self.inferencer_args.temperature,
+                    repetition_penalty=self.inferencer_args.repetition_penalty,
+                    do_sample=self.inferencer_args.do_sample,
+                )
 
             # only return the generation, trucating the input
             if self.model_args.arch_type != "vision_encoder_decoder":

src/lmflow/utils/flash_attention/llama_flash_attention.py

@@ -11,9 +11,9 @@
 
 #try to import flash_attn 2.x.x, if not, import flash_attn 1.x.x
 try:
-    from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func
+    from flash_attn.flash_attn_interface import flash_attn_func
 except:
-    from flash_attn.flash_attn_interface import flash_attn_unpadded_qkvpacked_func as flash_attn_varlen_qkvpacked_func
+    from flash_attn.flash_attn_interface import flash_attn_unpadded_qkvpacked_func as flash_attn_func
 
 from flash_attn.bert_padding import unpad_input, pad_input
 
@@ -22,136 +22,70 @@ def forward(
     self,
     hidden_states: torch.Tensor,
     attention_mask: Optional[torch.Tensor] = None,
-    position_ids: 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]]]:
-    """Input shape: Batch x Time x Channel
-
-    attention_mask: [bsz, q_len]
-    """
     bsz, q_len, _ = hidden_states.size()
 
-    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)
-    )
-    # [bsz, q_len, nh, hd]
-    # [bsz, nh, q_len, hd]
+    if self.config.pretraining_tp > 1:
+        raise ValueError("pretraining_tp > 1 is not supported for flash attention")
+    else:
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+    query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+    key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+    value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
 
     kv_seq_len = key_states.shape[-2]
-    # assert past_key_value is None, "past_key_value is not supported"
+
     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
-    )
-    # [bsz, nh, t, hd]
-    # assert not output_attentions, "output_attentions is not supported"
-    # assert not use_cache, "use_cache is not supported"
+    query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
     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)
 
     past_key_value = (key_states, value_states) if use_cache else None
 
-    # Flash attention codes from
-    # https://github.com/HazyResearch/flash-attention/blob/main/flash_attn/flash_attention.py
-
-    # transform the data into the format required by flash attention
-    # import pdb; pdb.set_trace()
-
-    if query_states.shape[-2] == 1 or query_states.shape[-2] != key_states.shape[-2]:
-        # decode token-by-token, do not use flash attention
-        # in incremental state, do not use flash attention
-        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):
-            raise ValueError(
-                f"Attention weights should be of size {(bsz * self.num_heads, q_len, kv_seq_len)}, but is"
-                f" {attn_weights.size()}"
-            )
-
-        if attention_mask is not 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()}"
-                )
-            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):
-            raise ValueError(
-                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
-                f" {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
-
-    # transform the data into the format required by flash attention
-    qkv = torch.stack(
-        [query_states, key_states, value_states], dim=2
-    )  # [bsz, nh, 3, q_len, hd]
-    qkv = qkv.transpose(1, 3)  # [bsz, q_len, 3, nh, hd]
-    # We have disabled _prepare_decoder_attention_mask in LlamaModel
-    # the attention_mask should be the same as the key_padding_mask
-    key_padding_mask = attention_mask
-
-    if key_padding_mask is None:
-        qkv = rearrange(qkv, "b s ... -> (b s) ...")
-        max_s = q_len
-        cu_q_lens = torch.arange(
-            0, (bsz + 1) * q_len, step=q_len, dtype=torch.int32, device=qkv.device
-        )
-        output = flash_attn_varlen_qkvpacked_func(
-            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
+    query_states, key_states, value_states = [
+        rearrange(x, "b h s d -> b s h d") for x in [query_states, key_states, value_states]
+    ]
+
+    input_dtype = query_states.dtype
+    if input_dtype == torch.float32:
+        # Handle the case where the model is quantized
+        if hasattr(self.config, "_pre_quantization_dtype"):
+            target_dtype = self.config._pre_quantization_dtype
+        else:
+            target_dtype = self.q_proj.weight.dtype
+
+        query_states = query_states.to(target_dtype)
+        key_states = key_states.to(target_dtype)
+        value_states = value_states.to(target_dtype)
+
+    # below output will have shape (batch_size, seqlen, nheads, headdim)
+    attn_output = flash_attn_func(query_states, key_states, value_states, causal=True)
+
+    if attn_output.size() != (bsz, q_len, self.num_heads, self.head_dim):
+        raise ValueError(
+            f"`attn_output` should be of size {(bsz, q_len, self.num_heads, self.head_dim)}, but is"
+            f" {attn_output.size()}"
         )
-        output = rearrange(output, "(b s) ... -> b s ...", b=bsz)
-    else:
-        nheads = qkv.shape[-2]
-        x = rearrange(qkv, "b s three h d -> b s (three h d)")
-        x_unpad, indices, cu_q_lens, max_s = unpad_input(x, key_padding_mask)
-        x_unpad = rearrange(
-            x_unpad, "nnz (three h d) -> nnz three h d", three=3, h=nheads
-        )
-        output_unpad = flash_attn_varlen_qkvpacked_func(
-            x_unpad, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
-        )
-        output = rearrange(
-            pad_input(
-                rearrange(output_unpad, "nnz h d -> nnz (h d)"), indices, bsz, q_len
-            ),
-            "b s (h d) -> b s h d",
-            h=nheads,
-        )
-    return self.o_proj(rearrange(output, "b s h d -> b s (h d)")), None, past_key_value
+
+    attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+    attn_output = self.o_proj(attn_output)
+    if output_attentions:
+        raise NotImplementedError("`output_attentions` is not supported when `use_flash_attn` is True")
+    attn_weights = None
+
+    return attn_output, attn_weights, past_key_value
 
 
 # Disable the transformation of the attention mask in LlamaModel as the flash attention