Support qwen2-7b with fused decoderlayer optimization on NPU #11912
+160
−45
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
plusbang
force-pushed
the
test-qwen2-7b
branch
from
3458b3a to
147f163
Compare
jason-dai
reviewed
Aug 26, 2024
Comment on lines
60
to
111
| elif model.config.model_type == "qwen2": | ||
| # for qwen2-1.5B and qwen2-7B |
There was a problem hiding this comment.
How about Qwen2-0.5B or 72B? Need to add check
There was a problem hiding this comment.
How about Qwen2-0.5B or 72B? Need to add check
Sure, have added related check : )
yangw1234
reviewed
Aug 26, 2024
python/llm/src/ipex_llm/transformers/npu_models/mp_models_base.py
Outdated
Show resolved
Hide resolved
plusbang
force-pushed
the
test-qwen2-7b
branch
from
83a2ddb to
ffdc410
Compare
plusbang
force-pushed
the
test-qwen2-7b
branch
from
c87ca0e to
10d6591
Compare
plusbang
changed the title
yangw1234
reviewed
Aug 28, 2024
| @@ -383,7 +383,7 @@ def update_cache(self, past_key_value, indexes): | |||
| self.load_cache_async() | |||
|
|
|||
| def load_cache_async(self): | |||
| self.load_wt_fn(len(self.input_ops), self._mm, self.kv_cache_c_handle) | |||
| self.load_wt_fn(len(self.input_ops), self._mm, self.kv_cache_c_handle, verify_size=True) | |||
There was a problem hiding this comment.
Not sure how much runtime overhead of verify_size, if it is large, maybe we can disable it at runtime and only use it when debugging?
There was a problem hiding this comment.
Not sure how much runtime overhead of
verify_size, if it is large, maybe we can disable it at runtime and only use it when debugging?
According to my experiment, the overhead seems quite small. BTW have removed it : )
yangw1234
approved these changes
Aug 28, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
Add qwen2-7B NPU support with fused decoderlayer and multi process optimization.
Details: https://github.com/analytics-zoo/nano/issues/1576#issuecomment-2314969138
Different from other model support, we use QuantizedLinear instead of FusedQwenLowBitDecoderlayer during prefill process.
2. User API changes
N/A
3. Summary of the change
4. How to test?
1234). And paste your action link here once it has been successfully finished.