Skip to content

Fixed self.k_proj.weight when using gptq #269

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

Merged
HolyFalafel merged 1 commit into from
Jun 25, 2024
Merged

Fixed self.k_proj.weight when using gptq #269

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
11 changes: 9 additions & 2 deletions optimum/habana/transformers/models/llama/modeling_llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -300,6 +300,13 @@ def get_k_proj_weight(self):
return self.k_proj.qweight
return self.k_proj.weight

def get_k_proj_weight_dtype(self):
""" 4bit quantization in GPTQ replaces the k_proj.weight with qweight.
Scales tensor gets the weight dtype. """
if hasattr(self.k_proj, 'qweight'):
return self.k_proj.scales.dtype
return self.k_proj.weight.dtype

def allocate_kv_cache(self, batch_size, max_seq_len, inp_seq_len):
cache_shape = (batch_size, self.num_key_value_heads, max_seq_len, self.head_dim)
device = self.get_k_proj_weight().device
Expand Down Expand Up @@ -418,9 +425,9 @@ def pre_attn_forward(
past_key_value = (self.k_cache.get_shape(), self.v_cache.get_shape())
else:
if past_key_value is None:
past_key = torch.zeros(key_states.shape, dtype=self.k_proj.weight.dtype, device=key_states.device)
past_key = torch.zeros(key_states.shape, dtype=self.get_k_proj_weight_dtype(), device=key_states.device)
past_value = torch.zeros(
key_states.shape, dtype=self.k_proj.weight.dtype, device=key_states.device
key_states.shape, dtype=self.get_k_proj_weight_dtype(), device=key_states.device
)
# Return list instead of tuple
past_key_value = [past_key, past_value]
Expand Down