Adam Ghanem

Hey all, Thanks for the tip on pinning. In the meantime, I did tinker around with `org-macs.el` and found that replacing the call to the`org-encode-time` wrapper function in `org-matcher-time` to...

Hello there to detail this blog, it is work by @Summer-Summer at [FSA-Lab](https://www.fsa-lab.org/) and others at Alibaba Research. The source code can be found at https://github.com/AlibabaResearch/flash-llm and https://github.com/usyd-fsalab/flash-llm. This work...

Hey there, although this does not directly solve the issue, I have trained the 1.3B model from scratch and the performance was quite decent, particularly with the EMA actor model....

Here is an example conversational output log from the EMA model: 

Hey @s-isaev no problem, this was trained on the base configurations for the 1.3B model provided on the Github. These are: - `training/step1_supervised_finetuning/training_scripts/single_node/run_1.3b.sh` - `training/step2_reward_model_finetuning/training_scripts/single_node/run_350m.sh` - `training/step3_rlhf_finetuning/training_scripts/single_node/run_1.3b.sh` The hardware configuration...

Hey there @feiliya333, for this model my commit ID for DeepSpeedExamples for that run was: `a5d4dc12828514ea7f427cd513665ba711dcd670`. EDIT: It seems I updated the repo on that machine in the meantime and...