deepspeedai/DeepSpeedExamples

View on GitHub

Example models using DeepSpeed

Stars: 6.8k
Forks: 1.1k
Open beginner issues: 0
Indexed issues: 300
Avg merge: 2d 16h
Merged PRs (30d): 1

Dominant language: Python
License: Apache-2.0
Last GitHub push: Aug 26, 2026
Latest indexed: Sep 20, 2026
Contributing guide: No contributing guide
Code of conduct: Code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

300 open issues indexed

Compute_rewards in PPO:rewards[j, start:ends[j]][-1] += reward_clip[j] is wrong Open

Difficulty 2/5 1-3 hours Newbie friendliness 52/100

deepspeedai/DeepSpeedExamples#996 · Opened Dec 22, 2025
Step1 failed when I run "bash training_scripts/opt/single_gpu/run_1.3b.sh" Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#995 · Opened Dec 16, 2025
One example, multiple config files Open

Difficulty 2/5 1-3 hours Newbie friendliness 52/100

deepspeedai/DeepSpeedExamples#989 · Opened Sep 8, 2025
moe example 404 Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#984 · Opened Jul 25, 2025
Possible to include an example of DeepNVMe + state dict Open

Difficulty 4/5 3-5 days Newbie friendliness 42/100

deepspeedai/DeepSpeedExamples#979 · 6 comments · Opened Jun 19, 2025
Apply Zero-3 and LoRA appears empty lora weight [0] Open

Difficulty 4/5 3-5 days Newbie friendliness 38/100

deepspeedai/DeepSpeedExamples#969 · 1 comment · Opened Apr 16, 2025
DeepSpeed-FastGen support ascend npu? Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

deepspeedai/DeepSpeedExamples#960 · 3 comments · Opened Mar 4, 2025
Why Does vf_loss Take the Maximum Value, Rendering Clamp Meaningless? Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#956 · 2 comments · Opened Feb 7, 2025
Is there any example about DeepSpeed Zero with Ulysses/Ulysses-offload Open

Difficulty 4/5 3-5 days Newbie friendliness 28/100

deepspeedai/DeepSpeedExamples#951 · Opened Jan 10, 2025
Domino + PP Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#950 · Opened Jan 9, 2025
Assertion `srcIndex < srcSelectDimSize` failed Open

deepspeedai/DeepSpeedExamples#946 · 1 comment · 1 assignee · Opened Dec 18, 2024
Question to attention computation Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#944 · Opened Dec 15, 2024
KV_cache offload Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#943 · 3 comments · 2 reactions · Opened Dec 15, 2024
How can I change the master_port when using deepspeed for multi-GPU on single node, i.e. localhost Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

deepspeedai/DeepSpeedExamples#936 · 4 comments · Opened Nov 20, 2024
Does DeepSpeed's Pipeline-Parallelism optimizer supports skip connections? Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#932 · Opened Oct 17, 2024
How to calculate training efficiency ,i.e tokens/sec of step 1 fine tuning of llama2 model ? Open

Difficulty 2/5 1-3 hours Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#923 · Opened Aug 29, 2024
Actor loss nan and Resizing model embedding Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#922 · 1 comment · Opened Aug 29, 2024
step2 without any response for a long time Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#915 · Opened Aug 15, 2024
Different zero stage the training memory compute Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#912 · Opened Jul 17, 2024
an error with gradient checkpointing in DeepspeedChat step 3 Open

deepspeedai/DeepSpeedExamples#908 · 1 assignee · Opened Jun 28, 2024
单机多卡进行RLHF在第三步中使用Qwen模型作Actor Model报错 Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#907 · 2 comments · Opened Jun 25, 2024
DeepSpeed-Chat step-1 hanging for a long time Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#906 · Opened Jun 19, 2024
CPU OOM when inferencing Llama3-70B-Chinese-Chat Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

deepspeedai/DeepSpeedExamples#904 · Opened May 20, 2024
cannot pickle 'Stream' object Open

Difficulty 3/5 1-2 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#903 · Opened May 17, 2024
can not run the test-gpt.sh because of assertionError Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#902 · Opened May 16, 2024
请问fastgen 是否支持长文本和序列并行推理 Open

Difficulty 5/5 Over a week Newbie friendliness 15/100

deepspeedai/DeepSpeedExamples#901 · Opened May 14, 2024
[Error] AutoTune: `connect to host localhost port 22: Connection refused` Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#894 · 1 reaction · Opened Apr 20, 2024
How to use deepspeed for multi-node and multi-card task in slurm cluster Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#893 · Opened Apr 18, 2024
Does Zero-Inference support TP? Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#892 · 11 comments · Opened Apr 16, 2024
Deepspeed support finetune extra model with lora ? Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#890 · 1 comment · Opened Apr 12, 2024
The actor constantly generates ['</s>'] or ['<|endoftext|></s>'] after 200 steps in RLHF with hybrid engine disabled Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#887 · 1 comment · Opened Apr 9, 2024
About multiple-thread attention computation on CPU using zero-inference example. Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#886 · Opened Apr 4, 2024
Suggested GPU to run the demo code of step2_reward_model_finetuning (DeepSpeed-Chat) Open

Difficulty 4/5 3-5 days Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#885 · Opened Apr 4, 2024
[REQUEST] More fine-grained distributed strategies for RLHF training Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#884 · Opened Apr 3, 2024
The reward value did not increase. Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#883 · 1 comment · Opened Apr 2, 2024
Confusion about Deepspeed Inference Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#879 · 1 comment · Opened Mar 25, 2024
`AttributeError: readonly attribute` while trying to run training/HelloDeepSpeed Open

Difficulty 3/5 1-2 days Newbie friendliness 35/100

deepspeedai/DeepSpeedExamples#878 · Opened Mar 21, 2024
[Bug] DeepSpeed Inference Does not Work with LLaMA (Latest verison) Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

deepspeedai/DeepSpeedExamples#867 · 3 comments · Opened Feb 29, 2024
[BUG in Stable Diffusion inference] There's an error on CUDAGraph when using deepspeed inference. How to fix it? Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

deepspeedai/DeepSpeedExamples#866 · 2 comments · Opened Feb 29, 2024
zero3 and enable hybrid engine are not suitable for llama2, how to solve it? Open

Difficulty 4/5 3-5 days Newbie friendliness 30/100

deepspeedai/DeepSpeedExamples#864 · 3 comments · Opened Feb 26, 2024
RLHF problems when using Qwen model Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#861 · 1 comment · Opened Feb 22, 2024
Codellama finetune Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#860 · Opened Feb 19, 2024
Throughput should be `num_queries/latency` as opposed to `num_clients/latency`? Open

Difficulty 3/5 1-2 days Newbie friendliness 38/100

deepspeedai/DeepSpeedExamples#858 · Opened Feb 3, 2024
The inaccurate flop results after several rounds Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#855 · 1 comment · Opened Jan 19, 2024
remove redundant code Open

Difficulty 1/5 Under an hour Newbie friendliness 45/100

deepspeedai/DeepSpeedExamples#852 · Opened Jan 13, 2024
Question: Why not padding to the same sequence length within the batch during the sft training phase? Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

deepspeedai/DeepSpeedExamples#849 · Opened Jan 12, 2024
torch.distributed.DistBackendError: NCCL error in: ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1333, remote process exited or there was a network error, NCCL version 2.18.6 Open

Difficulty 4/5 3-5 days Newbie friendliness 25/100

deepspeedai/DeepSpeedExamples#845 · 4 comments · Opened Jan 8, 2024
Invalidate trace cache @ step 0: expected module 0, but got module 6 Open

Difficulty 4/5 3-5 days Newbie friendliness 18/100

deepspeedai/DeepSpeedExamples#840 · Opened Dec 31, 2023
async_pipeline is not exposed in the library Open

Difficulty 1/5 Under an hour Newbie friendliness 48/100

deepspeedai/DeepSpeedExamples#835 · 1 comment · Opened Dec 20, 2023
Mistral and Orca Training Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

deepspeedai/DeepSpeedExamples#832 · Opened Dec 14, 2023

deepspeedai/DeepSpeedExamples

Get new issues in your inbox