Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / pacman100/llm-workshop issues and pull requests

#38 - [BUG]: Cannot run pipeline.py in personal_copilot folder

Issue - State: open - Opened by tahababou12 about 2 months ago

#36 - Update: Fixed directory name

Pull Request - State: open - Opened by suprateembanerjee 7 months ago

#35 - How to draw the loss graph?

Issue - State: open - Opened by allenliu88 8 months ago

#34 - Fine Tuning with LoRA failed during train step

Issue - State: open - Opened by arjun-mavonic 9 months ago - 1 comment

#31 - url in the readme is broken

Issue - State: closed - Opened by deven367 9 months ago - 1 comment

#30 - qwen moe support when using deepspeed

Pull Request - State: closed - Opened by pacman100 9 months ago

#29 - add loftq support

Pull Request - State: closed - Opened by pacman100 10 months ago

#28 - fsdp qlora and dsz3 qlora

Pull Request - State: closed - Opened by pacman100 11 months ago

#27 - fsdp qlora try

Pull Request - State: closed - Opened by pacman100 11 months ago

#26 - dsz3 qlora

Pull Request - State: closed - Opened by pacman100 11 months ago

#24 - Fix wrong fim token ids for deepseek-coder

Pull Request - State: closed - Opened by timxx 11 months ago

#23 - Add support for deepseek-coder base model and stable-code base model

Pull Request - State: closed - Opened by timxx 12 months ago

#22 - Packing = True

Issue - State: closed - Opened by palash04 12 months ago - 2 comments

#21 - FSDP+PEFT changes

Pull Request - State: closed - Opened by pacman100 12 months ago

#20 - Add data processing pipeline using HF datatrove library

Pull Request - State: closed - Opened by pacman100 12 months ago

#19 - Fix 4-bit quantization misspelling

Pull Request - State: closed - Opened by we1k about 1 year ago - 1 comment

#18 - Problem training with FSDP

Issue - State: open - Opened by agokrani about 1 year ago - 3 comments

#17 - Smangrul/lora fast train mode

Pull Request - State: closed - Opened by pacman100 about 1 year ago

#16 - Eval is like running forever

Issue - State: open - Opened by d5423197 about 1 year ago - 1 comment

#15 - chat assistant traning: CUDA out of memory

Issue - State: open - Opened by stevenhao about 1 year ago

#14 - use trl's `SFTTrainer`

Pull Request - State: closed - Opened by pacman100 about 1 year ago

#13 - remove the need to shard the model post saving when using FSDP

Pull Request - State: closed - Opened by pacman100 about 1 year ago

#12 - train Segmentation fault

Issue - State: open - Opened by bravelll about 1 year ago

#11 - Fixed usage of training command line args

Pull Request - State: closed - Opened by arpieb about 1 year ago

#10 - fix rng state for FIM training for code copilot training

Pull Request - State: closed - Opened by pacman100 about 1 year ago

#9 - Using device_map auto when launch acceleator

Issue - State: open - Opened by ronyadgar about 1 year ago

#8 - Delete unused args

Pull Request - State: closed - Opened by SingL3 over 1 year ago

#6 - Error on save_steps using FSDP

Issue - State: open - Opened by ghost over 1 year ago - 3 comments

#5 - Finetune 70B model on one node

Issue - State: open - Opened by yuanenming over 1 year ago

#5 - Finetune 70B model on one node

Issue - State: open - Opened by yuanenming over 1 year ago

#4 - Flash Attention for fine-tuning

Issue - State: closed - Opened by prince14322 over 1 year ago - 4 comments
Labels: solved

#3 - Update train.py

Pull Request - State: closed - Opened by pacman100 over 1 year ago

#2 - adding support for Flash Attn V2 for Falcon models

Pull Request - State: closed - Opened by pacman100 over 1 year ago

#1 - Incorrectness in Flash Attention

Issue - State: closed - Opened by mayank31398 over 1 year ago - 3 comments