VLM Pipeline (Intern,LLava)#256
Closed
qcdipankar wants to merge 2 commits intoquic:mainfrom
Closed
Conversation
Signed-off-by: Dipankar Sarkar <quic_dipankar@quicinc.com>
qcdipankar
requested review from
mohiso22,
quic-akuruvil,
quic-dhirajku and
vbaddi
quic-amitraj
requested changes
Feb 3, 2025
| return generate_func(**kwargs) | ||
|
|
||
| def generate_inputs_intern(self, **kwargs): | ||
| bs: int = constants.ONNX_EXPORT_EXAMPLE_BATCH_SIZE |
Contributor
There was a problem hiding this comment.
Please move this inside the modeling file.
| return inputs, dynamic_axes, output_names | ||
|
|
||
| def generate_inputs_llava(self, **kwargs): | ||
| bs: int = constants.ONNX_EXPORT_EXAMPLE_BATCH_SIZE |
Contributor
There was a problem hiding this comment.
Move this inside the modeling file modelling_llava.py
| # ) | ||
| # num_logits_to_keep = num_speculative_tokens + 1 | ||
| # if prefill_seq_len < num_logits_to_keep: | ||
| # raise ValueError( |
Contributor
There was a problem hiding this comment.
Remove commented lines.
| generation_len = self.ctx_len - input_len.max() # in standalone this is tensor | ||
| assert generation_len > 0, "generation length should be greater than zero" | ||
| generated_ids = np.full((batch_size, generation_len + 1), self.processor.tokenizer.pad_token_id) | ||
| # inputs["input_ids"]=torch.nn.functional.pad(inputs["input_ids"],(0,self.seq_len_constant-inputs["input_ids"].size(1)),"constant",self.pad_token_id) |
| MAX_RETRIES = 5 # This constant will be used set the maximum number of retry attempts for downloading a model using huggingface_hub snapshot_download | ||
| NUM_SPECULATIVE_TOKENS = 2 | ||
| CTX_LEN_VLM_LLAVA = 1280 | ||
| IMG_SIZE = 336 |
Contributor
There was a problem hiding this comment.
Are you using these at the time of export to define shapes?
|
|
||
|
|
||
| # if __name__ == "__main__": | ||
| # # model_name = "OpenGVLab/InternVL2_5-1B" |
Contributor
There was a problem hiding this comment.
Remove commented parts.
| if num_speculative_tokens: | ||
| compile_hash.update(to_hashable({"num_speculative_tokens": num_speculative_tokens})) | ||
|
|
||
| # import ipdb; ipdb.set_trace() |
| if hasattr(module, "__qeff_init__"): | ||
| module.__qeff_init__() | ||
| transformed = True | ||
|
|
Contributor
There was a problem hiding this comment.
Can we combine both if conditions?
| input_ids_size = input_ids.shape[1] | ||
| # attention_mask = inputs["attention_mask"] | ||
| inputs["input_ids"] = torch.nn.functional.pad( | ||
| inputs["input_ids"], (0, 3072 - input_ids_size), "constant", self.processor.tokenizer.pad_token_id |
Contributor
There was a problem hiding this comment.
Please avoid hardcoded value.
Contributor
There was a problem hiding this comment.
Make this value generic, and fetch from qpc session - prefill_seq_len. For whichever value it was compiled for.
| breakpoint() | ||
| self.model.config.use_cache = True | ||
| self.processor = processor | ||
| self.num_layers = model.config.text_config.num_hidden_layers |
Contributor
There was a problem hiding this comment.
Make fetching num_layers generic. Also the padding shape. Please refer llava PR, use a similar function, which fetches based on model architecture.
Contributor
|
Already addressed in #267. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Added Generic Framework to onboard and run VLMs in QEff