Ecosyste.ms: Issues
An open API service for providing issue and pull request metadata for open source projects.
GitHub / alipay/painlessinferenceacceleration issues and pull requests
#31 - Generation of Draft tokens and Trie tree creation
Issue -
State: closed - Opened by Vithulep about 2 months ago
#30 - 论文里看到Table 8 Inference Latency with Lookahead for vLLM.
Issue -
State: open - Opened by zjjznw123 3 months ago
#29 - 请问:batch_indices 是什么含义? LookaheadCache 的put、stream_put函数的最后一个参数ids是什么含义? class Tree(): 里面的idx代表什么含义?
Issue -
State: open - Opened by handsome-chips 4 months ago
#28 - stop_ids does not seem to be taking effect?
Issue -
State: open - Opened by jianyuheng 5 months ago
#27 - modeling_qwen attention not use multi branch position ids & attention_mask
Issue -
State: open - Opened by snippetzero 6 months ago
- 1 comment
#26 - lookahead with do_sample=True does not take temperature, top_k, top_p
Issue -
State: open - Opened by learning-chip 7 months ago
- 2 comments
#25 - How is verification done in PAIN?
Issue -
State: open - Opened by jivanph 8 months ago
#24 - Do lookahead and repetition_penalty conflict?
Issue -
State: open - Opened by zhanweiw 8 months ago
- 1 comment
#23 - AntRAG
Issue -
State: open - Opened by nrmer 8 months ago
- 1 comment
#22 - Changing naive attention to SDPA gives wrong result for batched llama example
Issue -
State: open - Opened by learning-chip 9 months ago
- 3 comments
#21 - size of memory footprint
Issue -
State: closed - Opened by nrmer 9 months ago
- 1 comment
#20 - 是否支持Qwen 1.5?
Issue -
State: open - Opened by hwang824 9 months ago
#19 - TODO in PainlessInferenceAcceleration/pia/lookahead/common/lookahead_cache.py
Issue -
State: closed - Opened by nrmer 9 months ago
- 1 comment
#18 - Clarification on edls/dls/ft in perf_check
Issue -
State: closed - Opened by nrmer 9 months ago
- 1 comment
#17 - when the batchSzie > 1, the lookahead not work.
Issue -
State: closed - Opened by yuenyu1 9 months ago
#17 - when the batchSzie > 1, the lookahead not work.
Issue -
State: closed - Opened by yuenyu1 9 months ago
#16 - Error: no attribute 'rope_theta'` for llama2 model
Issue -
State: closed - Opened by learning-chip 9 months ago
- 1 comment
#15 - Consultation on Trie Tree Maintenance?
Issue -
State: closed - Opened by ZipECHO 10 months ago
- 7 comments
#14 - Counting how many forward passes/steps were done when using PAIN
Issue -
State: open - Opened by jivanph 10 months ago
- 4 comments
#13 - Consider Support CodeLlama?
Issue -
State: open - Opened by RainYQ 10 months ago
- 2 comments
#12 - 是不支持ptuning之后的模型吗?
Issue -
State: closed - Opened by 13269279918 10 months ago
- 1 comment
#11 - 为什么我测试没有性能提升?
Issue -
State: closed - Opened by MeJerry215 10 months ago
- 3 comments
#10 - Dockerfile集成pia无法使用
Issue -
State: closed - Opened by May-Yaha 10 months ago
- 7 comments
#9 - In the benchmark studies, how are the draft tokens generated?
Issue -
State: open - Opened by jivanph 10 months ago
- 9 comments
#8 - Update README.md
Pull Request -
State: closed - Opened by eltociear 10 months ago
#7 - BUG for chatglm3-6b and Qwen-14B-Int4
Issue -
State: closed - Opened by AGI-Jarvis 10 months ago
- 7 comments
#6 - 为什么感觉没有什么效果
Issue -
State: closed - Opened by shinerdeng 10 months ago
- 6 comments
#5 - 速度确实有提升,但是生成的质量存在问题
Issue -
State: closed - Opened by dafen12 10 months ago
- 2 comments
#4 - 这个lookahead的实现和原版hao-ailab的实现相比优化点在哪里?
Issue -
State: closed - Opened by janelu9 10 months ago
- 1 comment
#3 - How the performance VS vLLM inference(vLLM vs Lookahead)
Issue -
State: open - Opened by buptygz 10 months ago
- 5 comments
#3 - How the performance VS vLLM inference(vLLM vs Lookahead)
Issue -
State: open - Opened by buptygz 10 months ago
- 5 comments
#2 - 可以批(batch)推理吗?比如一次推理256个输入?满级压榨GPU!^_^
Issue -
State: closed - Opened by janelu9 10 months ago
- 3 comments
#1 - 请问有支持量化版本的qwen架构的计划吗?
Issue -
State: closed - Opened by xiningnlp 10 months ago
- 2 comments