Skip to content

Address Phi modeling update #2383

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
kausky wants to merge 6 commits into from
Closed

Address Phi modeling update #2383

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
ad3dcdc
fix: update layer_norm_epsilon in phi_1_5 tp layer_norm_eps
kausky Jan 8, 2024
cd9a6ef
fix(phi-1_5): layer_norm_eps and layer_norm_epsilon checking.
kausky Jan 8, 2024
2803c81
fix config variable names
simon-mo Jan 12, 2024
0b643db
rename phi
simon-mo Jan 12, 2024
37bd3e6
Merge branch 'main' of github.com:vllm-project/vllm into phi-1.5-fix
simon-mo Jan 12, 2024
6d73b82
reset debugging
simon-mo Jan 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 16 additions & 6 deletions vllm/model_executor/models/phi.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -113,13 +113,17 @@ def __init__(self,
)

scaling = self.head_size**-0.5
rotary_dim = config.rotary_dim

# https://huggingface.co/microsoft/phi-2/blob/cb2f4533604d8b67de604e7df03bfe6f3ca22869/modeling_phi.py#L278-L281
assert config.rope_scaling is None, "rope_scaling is not supported"
head_dim = config.hidden_size // config.num_attention_heads
rotary_dim = int(config.partial_rotary_factor * head_dim)
assert rotary_dim % 2 == 0

# pylint: disable=C0301
# Refer to:
# https://huggingface.co/microsoft/phi-1_5/blob/d212a789620c380ff32ca1d1ee9943a777360987/modeling_phi.py#L518
rope_theta = 10000
rope_theta = getattr(config, "rope_theta", 10000)
max_position_embeddings = getattr(config, "n_positions", 2048)
self.rotary_emb = get_rope(
self.head_size,
Expand Down Expand Up @@ -166,7 +170,7 @@ def __init__(self,
linear_method=linear_method,
)
quant_config = getattr(linear_method, "quant_config", None)
self.act = get_act_fn(config.activation_function, quant_config,
self.act = get_act_fn(config.hidden_act, quant_config,
n_inner)

def forward(self, hidden_states):
Expand All @@ -175,6 +179,12 @@ def forward(self, hidden_states):
hidden_states, _ = self.fc2(hidden_states)
return hidden_states

def _get_layer_norm_eps(config: PretrainedConfig) -> float:
# check for layer_norm_eps in case of phi-1.5 and layer_norm_epsilon for phi-2
layer_norm_eps = getattr(config, "layer_norm_eps", None)
if layer_norm_eps is None:
layer_norm_eps = getattr(config, "layer_norm_epsilon", 1e-6)
return layer_norm_eps

class PhiLayer(nn.Module):

Expand All @@ -183,7 +193,7 @@ def __init__(self,
linear_method: Optional[LinearMethodBase] = None):
super().__init__()
self.ln = nn.LayerNorm(config.hidden_size,
eps=config.layer_norm_epsilon)
eps=_get_layer_norm_eps(config))
self.mixer = PhiAttention(config, linear_method)
self.mlp = PhiMLP(config, linear_method)

Expand Down Expand Up @@ -245,7 +255,7 @@ class PhiCausalLMHead(nn.Module):
def __init__(self, config: PretrainedConfig):
super().__init__()
self.ln = nn.LayerNorm(config.hidden_size,
eps=config.layer_norm_epsilon)
eps=_get_layer_norm_eps(config))
self.linear = ParallelLMHead(config.vocab_size,
config.hidden_size,
bias=True)
Expand Down Expand Up @@ -304,4 +314,4 @@ def load_weights(self,
param = params_dict[name]
weight_loader = getattr(param, "weight_loader",
default_weight_loader)
weight_loader(param, loaded_weight)
weight_loader(param, loaded_weight)