Merge pull request #2319 from ftgreat/main

Add patch_mixtral_replace_moe_impl for full training Mitral using DeepSpeed Zero3
This commit is contained in:
hoshi-hiyouga 2024-01-24 15:32:26 +08:00 committed by GitHub
commit 51ad35b3c7
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 40 additions and 0 deletions

View File

@ -268,6 +268,43 @@ def patch_config(
_configure_quantization(config, tokenizer, model_args, config_kwargs)
def patch_mixtral_replace_moe_impl() -> None:
import torch.nn.functional as F
def mlp_forward(self, hidden_states):
current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(hidden_states)
current_hidden_states = self.w2(current_hidden_states)
return current_hidden_states
## Ref. https://huggingface.co/deepseek-ai/deepseek-moe-16b-base/blob/main/modeling_deepseek.py
def moe_forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
batch_size, sequence_length, hidden_dim = hidden_states.shape
hidden_states = hidden_states.view(-1, hidden_dim)
# router_logits: (batch * sequence_length, n_experts)
router_logits = self.gate(hidden_states)
routing_weights = F.softmax(router_logits, dim=1, dtype=torch.float)
topk_weight, topk_idx = torch.topk(routing_weights, self.top_k, dim=-1, sorted=False)
topk_weight /= topk_weight.sum(dim=-1, keepdim=True)
# we cast back to the input dtype
topk_weight = topk_weight.to(hidden_states.dtype)
hidden_states = hidden_states.repeat_interleave(self.top_k, dim=0)
y = torch.empty_like(hidden_states)
flat_topk_idx = topk_idx.view(-1)
for i in range(self.num_experts):
expert = self.experts[i]
y[flat_topk_idx == i] = expert(hidden_states[flat_topk_idx == i])
y = (y.view(*topk_weight.shape, -1) * topk_weight.unsqueeze(-1)).sum(dim=1)
final_hidden_states = y.reshape(batch_size, sequence_length, hidden_dim)
return final_hidden_states, router_logits
from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
from transformers.models.mixtral.modeling_mixtral import MixtralBLockSparseTop2MLP
MixtralBLockSparseTop2MLP.forward = mlp_forward
MixtralSparseMoeBlock.forward = moe_forward
def patch_model(
model: "PreTrainedModel", tokenizer: "PreTrainedTokenizer", model_args: "ModelArguments", is_trainable: bool
) -> None:
@ -290,6 +327,9 @@ def patch_model(
from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
set_z3_leaf_modules(model, [MixtralSparseMoeBlock])
if is_trainable:
patch_mixtral_replace_moe_impl()
def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") -> None:
def tie_weights(self: "AutoModelForCausalLMWithValueHead") -> None: