Fix graph breaks in torch compile mode#52
Closed
scsudhak-intel wants to merge 18 commits into
Closed
Conversation
* Expose Llama Fused OPs control from run_lora_clm.py * Update as per review comments
* enable internal kv bucket in llama * initialize bucket_internal for CI * make bucket_internal more clear * further perf optim while max length is not multiple of bucket size
* [SW-173358] add first token prints * [SW-173358] rename x to outputs * [SW-173358] make style
* Enable Flash Attention in recompute and causal modes * Add flash_attention_causal_mask to generation utils * Propagate Flash Attention causal_mask to finetuning example * Modify README example and provide additional description * Add flash_attention_causal_mask to FT README
* enable loading falcon-180b ckpt in .safetensors format * Address comments borrowing transformer's way of reading ckpt file * address comments
mark_step() should not be called for eager mode Signed-off-by: Manoj Kumar <mkumar@habana.ai>
Co-authored-by: Sun Choi <schoi@habana.ai>
* enable loading falcon-180b ckpt in .safetensors format * Address comments borrowing transformer's way of reading ckpt file * address comments * Update ckpt loading PR#15 reads a set of ckpt file names from the index json file. When OH downloads files from the hub instead of loading from a cache dir, get_repo_root() skips downloading the index json file. Thus the PR#15 fails to load file names. This PR scans the path and returns a list of names that matches the pattern * import modeling_utils from transformers
Co-authored-by: Sayantan Sarkar <supersarkar@gmail.com>
* Further fixes for performance with internal bucketing. Also add clear cache() to save memory. make style changes also added. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Calculate kv cache sliding idx for the decode phase only. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Add hpu graphs check for clear cache. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
* Adding a flag whether to save checkpoint or not * Add the flag to a model run script
* [SW-174850] Fix for Nonetype in image * Using media external reader API * minor fixes * output info fix * Update media reader function name * make style
Signed-off-by: Sanju C Sudhakaran <scsudhakaran@habana.ai>
Signed-off-by: Sanju C Sudhakaran <scsudhakaran@habana.ai>
Signed-off-by: Sanju C Sudhakaran <scsudhakaran@habana.ai>
bhargaveede
force-pushed
the
habana-main
branch
from
8d30377 to
4cf8089
Compare
astachowiczhabana
pushed a commit
that referenced
this pull request
Jan 23, 2025
* Initial commit => add overrides to support bnb on HPU * Change quantizer file name * Added HPU specific checks and updates (#52) * supports hpu nf4 quant/dequant * added tranformer/quantizers to bitsandbytes * Deleted transformer/quantizers * Fix BnB inference (#60) * update inference test (#63) * Added 4-bit training script (#69) * adapted test files with framework * Adjusted test files format --------- Co-authored-by: Vivek <vgoel@habana.ai>
xinyu-intel
pushed a commit
that referenced
this pull request
Mar 4, 2025
* Initial commit => add overrides to support bnb on HPU * Change quantizer file name * Added HPU specific checks and updates (#52) * supports hpu nf4 quant/dequant * added tranformer/quantizers to bitsandbytes * Deleted transformer/quantizers * Fix BnB inference (#60) * update inference test (#63) * Added 4-bit training script (#69) * adapted test files with framework * Adjusted test files format --------- Co-authored-by: Vivek <vgoel@habana.ai>
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
10 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.
Training script uses dynamic control flow based on the success of a module import, which caused a graph break.
This patch handles the control flow through boolean flags which avoids graph breaks.