Search results

Degraded Performance Converting Transformer Engine-Trained LM to HuggingFace

bugSomething isn't workingSomething isn't working

Status: Open.

#1353 In EleutherAI/gpt-neox;

· Kyle1668 opened on Apr 16, 2025

Unable to Use SwiGLU with Transformer Engine

bugSomething isn't workingSomething isn't working

Status: Open.

#1347 In EleutherAI/gpt-neox;

· Kyle1668 opened on Mar 13, 2025

AssertionError in position embedding (potentially due to missing <code>clear_cache</code> between batches of data)

bugSomething isn't workingSomething isn't working

Status: Open.

#1340 In EleutherAI/gpt-neox;

· ameyagodbole opened on Feb 25, 2025

convert_hf_llama_to_neox.py appears incompatible with LLAMA-3.1-70B

bugSomething isn't workingSomething isn't working

Status: Open.

#1337 In EleutherAI/gpt-neox;

· tijmen opened on Feb 14, 2025

DeeperSpeed-Pytorch Incompatibility

bugSomething isn't workingSomething isn't working

Status: Open.

#1335 In EleutherAI/gpt-neox;

· tijmen opened on Feb 5, 2025

How are multiple datasets loaded?

question

Status: Open.

#1330 In EleutherAI/gpt-neox;

· fxnie opened on Dec 17, 2024

[Questions] Context length extension.

question

Status: Open.

#1327 In EleutherAI/gpt-neox;

· iPRET opened on Dec 16, 2024

How to set the learning rate using WSD? As mentioned in MiniCPM

feature requestNew feature or requestNew feature or request

Status: Open.

#1326 In EleutherAI/gpt-neox;

· fxnie opened on Dec 16, 2024

Error when converting sequential model to HF

bugSomething isn't workingSomething isn't working

Status: Open.

#1323 In EleutherAI/gpt-neox;

· SilverSulfide opened on Nov 20, 2024

Can <code>preprocess_data.py</code> support Huggingface Dataset?

feature requestNew feature or requestNew feature or request

Status: Open.

#1321 In EleutherAI/gpt-neox;

· cafeii opened on Nov 13, 2024

_forward_step_fn does not always return two values so eval.py breaks if is_pipe_parallel is false

bugSomething isn't workingSomething isn't working

Status: Open.

#1320 In EleutherAI/gpt-neox;

· markNZed opened on Nov 12, 2024

DeeperSpeed cannot support BFloat16 and PipelineParallelism

bugSomething isn't workingSomething isn't working

Status: Open.

#1307 In EleutherAI/gpt-neox;

· jahatef opened on Oct 15, 2024