Use KV cache till input seq len for prefill phase #154
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
Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
puneeshkhanna
requested review from
ssarkar2,
bhargaveede,
vivekgoe,
mandy-li,
libinta and
dvarshney-habana
as code owners
|
Updated command (remove --reuse_cache , setting PT_HPUGRAPH_DISABLE_TENSOR_CACHE=1 automatically taken care) python ../gaudi_spawn.py --use_deepspeed --world_size 8 run_generation.py --model_name_or_path /mnt/weka/data/pytorch/llama2/Llama-2-70b-hf/ --use_hpu_graphs --use_kv_cache --max_input_tokens 2048 --max_new_tokens 2048 --batch_size 200 --attn_softmax_bf16 --trim_logits --bf16 --warmup 2 --n_iterations 2 --limit_hpu_graphs --bucket_internal --bucket_size 128 Also requires pytorch-integration patch - https://gerrit.habana-labs.com/#/c/408363/ |
dvarshney-habana
approved these changes
Apr 11, 2024
sushildubey171
pushed a commit
that referenced
this pull request
Apr 12, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
astachowiczhabana
pushed a commit
that referenced
this pull request
Apr 19, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
astachowiczhabana
pushed a commit
that referenced
this pull request
Apr 22, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
astachowiczhabana
pushed a commit
that referenced
this pull request
Apr 24, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
astachowiczhabana
pushed a commit
that referenced
this pull request
Apr 24, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
puneeshkhanna
pushed a commit
to puneeshkhanna/optimum-habana-fork
that referenced
this pull request
May 2, 2024
* Use KV cache till input seq len for prefill phase. Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable. All the changes are protected by bucket internal flag. Signed-off-by: Puneesh Khanna <pkhanna@habana.ai> * Revert initialization of KV cache * Set PT_HPUGRAPH_DISABLE_TENSOR_CACHE flag * remove os import * remove commented print --------- Signed-off-by: Puneesh Khanna <pkhanna@habana.ai>
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.
Pad KV cache to full input + new tokens len for decode phase. Delete the KV cache used as inputs by HPU graphs after full prompt generation. Ensure KV cache is not returned as output tensor during decode phase. Deletion of KV cache input tensor used by HPU graphs needs to be protected by PT_HPUGRAPH_DISABLE_TENSOR_CACHE env variable.
All the changes are protected by bucket internal flag.
What does this PR do?
Fixes # (issue)
Before submitting