Skip to content

[quantization] Pass static data as inputs #537

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
mhs4670go merged 1 commit into from
Mar 10, 2026
Merged

[quantization] Pass static data as inputs #537

Show file tree
Hide file tree
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
25 changes: 21 additions & 4 deletions tico/quantization/wrapq/examples/llama/quantize_decoder_layer_prefill.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@

MODEL_NAME = "Maykeye/TinyLLama-v0"
MAX_SEQ = 256
static_data_inside_the_layer = True

model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, dtype="float32")
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, legacy=False)
Expand Down Expand Up @@ -139,10 +140,26 @@ def make_fixed_inputs(prompt: str):
example_hidden = torch.randn(B, S, D)

with SuppressWarning(UserWarning, ".*"):
cm = tico.convert(
qlayer,
(example_hidden,),
)
if static_data_inside_the_layer is True:
cm = tico.convert(
qlayer,
(example_hidden,),
)
else:
qattn = qlayer.wrapped
attn_mask = qattn._slice_causal(S, "cpu").squeeze(0)
dtype = example_hidden.dtype
pos_embeds = (
qattn.rope_cos_template.cpu().to(dtype),
qattn.rope_sin_template.cpu().to(dtype),
)

cm = tico.convert(
qlayer,
(example_hidden,),
kwargs={"attention_mask": attn_mask, "position_embeddings": pos_embeds},
)

# Note that the model is not fully quantized.
cm.save(save_path)

Expand Down
16 changes: 8 additions & 8 deletions tico/quantization/wrapq/wrappers/llama/quant_decoder_layer_prefill.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -193,9 +193,9 @@ def forward(
L = hidden_states.size(1)
attention_mask = self._slice_causal(L, hidden_states.device)
attention_mask = attention_mask.squeeze(0)
attention_mask = self._fq(
attention_mask, self.obs_causal_mask
) # let it be quantized immediately
attention_mask = self._fq(
attention_mask, self.obs_causal_mask
) # let it be quantized immediately

if position_embeddings is None:
position_embeddings = (
Expand All @@ -206,11 +206,11 @@ def forward(
dtype=hidden_states.dtype, device=hidden_states.device
),
)
cos, sin = position_embeddings
position_embeddings = (
self._fq(cos, self.obs_cos),
self._fq(sin, self.obs_sin),
)
cos, sin = position_embeddings
position_embeddings = (
self._fq(cos, self.obs_cos),
self._fq(sin, self.obs_sin),
)

attn_out = self.self_attn(
hidden_states=hidden_states,
Expand Down