github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-07-26 13:01:21 +00:00
Code Issues Packages Projects Releases Wiki Activity

implement memory efficient logprob

Browse Source
This commit is contained in:
YeAnbang 2025-06-18 10:24:48 +00:00
parent 30a6859f77
commit e3d56cbd86
2 changed files with 43 additions and 48 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
applications/ColossalChat/coati/distributed/grpo_consumer.py
View File

@ -6,7 +6,7 @@ import torch
import wandb import wandb
from coati.distributed.consumer import BaseConsumer from coati.distributed.consumer import BaseConsumer
from coati.distributed.loss import PolicyLoss from coati.distributed.loss import PolicyLoss
from coati.distributed.utils import calc_action_log_probs from coati.distributed.utils import memory_efficient_logprob
from coati.trainer.utils import all_reduce_mean, all_reduce_sum from coati.trainer.utils import all_reduce_mean, all_reduce_sum
from transformers import AutoModelForCausalLM, AutoTokenizer from transformers import AutoModelForCausalLM, AutoTokenizer
@ -280,21 +280,12 @@ class GRPOConsumer(BaseConsumer):
) )
if self.booster.plugin.stage_manager.is_last_stage(): if self.booster.plugin.stage_manager.is_last_stage():
reference_action_log_probs = torch.zeros( reference_action_log_probs = memory_efficient_logprob(
(input_ids_forward_micro_batch.size(0), num_action), reference_model_outputs["outputs"]["logits"],
device=input_ids_forward_micro_batch.device, input_ids_forward_micro_batch,
num_action,
shard_config=self.plugin.shard_config,
) )
for i in range(reference_action_log_probs.size(0)):
# activation for log_softmax is too large if vocab size and sequence length are large
# e.g., when using 152064 vocab size with 32K seqence length and a micro batch size of 4 (for pp=4 for example),
# this activation sorely takes 152064*32000*4*4/1024/1024/1024=72.5GB
reference_action_log_probs[i, :] += calc_action_log_probs(
reference_model_outputs["outputs"]["logits"][i : i + 1]
/ self.generate_config["temperature"],
input_ids_forward_micro_batch[i : i + 1],
num_action,
self.plugin.shard_config,
)[0]
else: else:
# Dummy reference logprobs for data iterator. # Dummy reference logprobs for data iterator.
reference_action_log_probs = None reference_action_log_probs = None
@ -316,19 +307,12 @@ class GRPOConsumer(BaseConsumer):
def _criterion(outputs, inputs): def _criterion(outputs, inputs):
action_logits = outputs.logits action_logits = outputs.logits
action_log_probs = torch.zeros( action_log_probs = memory_efficient_logprob(
(inputs["input_ids"].size(0), num_action), device=action_logits.device action_logits,
inputs["input_ids"],
num_action,
shard_config=self.plugin.shard_config,
) )
for i in range(action_log_probs.size(0)):
# activation for log_softmax is too large if vocab size and sequence length are large
# e.g., when using 152064 vocab size with 32K seqence length and a micro batch size of 4 (for pp=4 for example),
# this activation sorely takes 152064*32000*4*4/1024/1024/1024=72.5GB
action_log_probs[i, :] += calc_action_log_probs(
action_logits[i : i + 1] / self.generate_config["temperature"],
inputs["input_ids"][i : i + 1],
num_action,
self.plugin.shard_config,
)[0]
if "reference_action_log_probs" in inputs: if "reference_action_log_probs" in inputs:
per_token_kl = ( per_token_kl = (
torch.exp(inputs["reference_action_log_probs"] - action_log_probs) torch.exp(inputs["reference_action_log_probs"] - action_log_probs)
@ -370,16 +354,15 @@ class GRPOConsumer(BaseConsumer):
mean_kl.append(kl) mean_kl.append(kl)
mean_loss.append(all_reduce_mean(loss, self.plugin).data) mean_loss.append(all_reduce_mean(loss, self.plugin).data)
else: else:
policy_model_logits = self.policy_model( policy_model_logits = self.policy_model(
input_ids=input_ids_forward_micro_batch, input_ids=input_ids_forward_micro_batch,
attention_mask=attention_mask_forward_micro_batch, attention_mask=attention_mask_forward_micro_batch,
).logits ).logits
action_log_probs = calc_action_log_probs( action_log_probs = memory_efficient_logprob(
policy_model_logits / self.generate_config["temperature"], policy_model_logits / self.generate_config["temperature"],
input_ids_forward_micro_batch, input_ids_forward_micro_batch,
num_action, num_action,
self.plugin.shard_config, shard_config=self.plugin.shard_config,
) )
if self.policy_loss_fn.beta > 0: if self.policy_loss_fn.beta > 0:
@ -388,7 +371,7 @@ class GRPOConsumer(BaseConsumer):
input_ids=input_ids_forward_micro_batch, input_ids=input_ids_forward_micro_batch,
attention_mask=attention_mask_forward_micro_batch, attention_mask=attention_mask_forward_micro_batch,
).logits ).logits
reference_action_log_probs = calc_action_log_probs( reference_action_log_probs = memory_efficient_logprob(
reference_model_logits / self.generate_config["temperature"], reference_model_logits / self.generate_config["temperature"],
input_ids_forward_micro_batch, input_ids_forward_micro_batch,
num_action, num_action,

46
applications/ColossalChat/coati/distributed/utils.py
View File

@ -71,31 +71,43 @@ def log_probs_from_logits(logits: torch.Tensor, labels: torch.Tensor) -> torch.T
return per_label_logps.squeeze(-1) return per_label_logps.squeeze(-1)
def calc_action_log_probs( def memory_efficient_logprob(
logits: torch.Tensor, logits: torch.Tensor,
sequences: torch.LongTensor, inputs: torch.Tensor,
num_actions: int, num_action: int,
shard_config, chunk_size: int = 2048,
shard_config: Any = None,
vocab_size: int = None, vocab_size: int = None,
) -> torch.Tensor: ) -> torch.Tensor:
"""Calculate action log probs. """
Calculate action log probs in a memory-efficient way by processing in chunks.
Args: Args:
logits (torch.Tensor): Output tensor of Actor.forward.logits. logits (torch.Tensor): Output tensor of Actor.forward.logits.
sequences (torch.LongTensor): Input sequences. inputs (torch.LongTensor): Input sequences.
num_actions (int): Number of actions. num_action (int): Number of actions.
shard_config chunk_size (int, optional): Size of each chunk to process. Default is 2048.
vocab_size shard_config: Shard configuration for distributed computation.
vocab_size (int, optional): Vocabulary size. Default is None.
Returns: Returns:
torch.Tensor: Action log probs. torch.Tensor: Action log probs.
""" """
# labels: torch.Tensor, # [B, S] or [B, S, Vocab_size] action_log_probs = torch.zeros((logits.size(0), num_action), device=logits.device, dtype=logits.dtype)
# logits: torch.Tensor, # [B, S, Vocab_size] context_length = logits.size(1) - num_action
log_probs = dist_log_prob(sequences, logits, shard_config, vocab_size, logits.dtype) for i in range(action_log_probs.size(0)):
log_probs = log_probs.squeeze(-1) # loop over each sample in the micro-batch
return log_probs[:, -num_actions:] for start in range(context_length, logits.size(1), chunk_size):
end = min(start + chunk_size, logits.size(1))
# calculate log probs in chunks to save memory
log_probs = dist_log_prob(
inputs[i : i + 1, start - 1 : end],
logits[i : i + 1, start - 1 : end],
shard_config,
vocab_size,
logits.dtype,
) # [1, chunk_size, 1]
log_probs = log_probs.squeeze(-1)
action_log_probs[i, start - context_length : end - context_length] += log_probs[0]
return action_log_probs
def masked_mean(tensor: torch.Tensor, mask: torch.Tensor, dim: int = 1) -> torch.Tensor: def masked_mean(tensor: torch.Tensor, mask: torch.Tensor, dim: int = 1) -> torch.Tensor: