Skip to content
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.

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

fix a bug: Convert multi weight slice file #60

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open
Changes from all commits
Commits
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
13 changes: 8 additions & 5 deletions distserve/downloader/converter.py
Original file line number Diff line number Diff line change
Expand Up @@ -114,13 +114,15 @@ def preprocess_llama2(tensor_dict: Dict[str, torch.Tensor])\
V_WEIGHT = PREFIX + "layers.{0}.self_attn.v_proj.weight"
O_WEIGHT = PREFIX + "layers.{0}.self_attn.o_proj.weight"

num_layers = max(int(regex.findall(x)[0]) for x in filter(regex.match, tensor_dict)) + 1

end_layers = max(int(regex.findall(x)[0]) for x in filter(regex.match, tensor_dict)) + 1
beg_layers = min(int(regex.findall(x)[0]) for x in filter(regex.match, tensor_dict))

head_dim = 128
num_q_heads = tensor_dict[Q_WEIGHT.format(0)].size(0) // head_dim
num_q_heads = tensor_dict[Q_WEIGHT.format(24)].size(0) // head_dim

# Coallesce wq, qk, qv into one tensor, layers.{i}.attention.wqkv.weight
for i in range(num_layers):

for i in range(beg_layers,end_layers):
q = tensor_dict[Q_WEIGHT.format(i)].T # [hidden_size, num_q_heads*head_dim]
k = tensor_dict[K_WEIGHT.format(i)].T # [hidden_size, num_kv_heads*head_dim]
v = tensor_dict[V_WEIGHT.format(i)].T # [hidden_size, num_kv_heads*head_dim]
Expand All @@ -131,7 +133,7 @@ def preprocess_llama2(tensor_dict: Dict[str, torch.Tensor])\
del tensor_dict[V_WEIGHT.format(i)]

# Transpose wo
for i in range(num_layers):
for i in range(beg_layers,end_layers):
tensor_dict[O_WEIGHT.format(i)] = \
tensor_dict[O_WEIGHT.format(i)].T.contiguous() # [num_q_heads*head_dim, hidden_size]

Expand Down Expand Up @@ -436,6 +438,7 @@ def convert_weights(
# Preprocess
print("Preprocessing")
preprocessor = PREPROCESSOR[model]
print(state_dict)
tensor_dict, num_q_heads, head_dim = preprocessor(state_dict)

# The final step: divide the weights and save them to files
Expand Down