Search results

Memory Leak when initialising Qwen model training with multihost runner

bugSomething isn't workingSomething isn't working

Status: Open.

#2374 In AI-Hypercomputer/maxtext;

· jedcheng opened on Sep 20, 2025

[feature request] Make z_loss factor configurable

feature request

Status: Open.

#2352 In AI-Hypercomputer/maxtext;

· bzantium opened on Sep 16, 2025

[feature request] Support chunking (splitting) long sequences instead of truncation during tokenization

feature request

Status: Open.

#2344 In AI-Hypercomputer/maxtext;

· bzantium opened on Sep 15, 2025

[feature request] Support of GLM4.5V and Deepseek V3.1

feature request

Status: Open.

#2330 In AI-Hypercomputer/maxtext;

· yuanfz98 opened on Sep 11, 2025

QAT SFT Training Fails with Qwen3 MOE

bugSomething isn't workingSomething isn't working

Status: Open.

#2305 In AI-Hypercomputer/maxtext;

· chaserogo opened on Sep 7, 2025

Failed build: Tests

build failed

Status: Open.

#2233 In AI-Hypercomputer/maxtext;

· github-actions opened on Aug 24, 2025

Outdated <code>constraints_gpu.txt</code>

bugSomething isn't workingSomething isn't working

Status: Open.

#2212 In AI-Hypercomputer/maxtext;

· emergenz opened on Aug 20, 2025

Quality-checking tooling: let's move to newer ones, pytype is deprecated, <code>ruff</code> is popular, etc.

feature request

Status: Open.

#2210 In AI-Hypercomputer/maxtext;

· SamuelMarks opened on Aug 20, 2025

Run pretraining of llama 7b/8b models on GPUs using a3/1vm.sh

bugSomething isn't workingSomething isn't working

Status: Open.

#2193 In AI-Hypercomputer/maxtext;

· giacomoni opened on Aug 15, 2025

Gemma3 Model Token Logits Mismatch with Different max_prefill_predict_length

bugSomething isn't workingSomething isn't working

Status: Open.

#2179 In AI-Hypercomputer/maxtext;

· babyplutokurt opened on Aug 14, 2025

Conversion to unscanned format failed

bugSomething isn't workingSomething isn't working

Status: Open.

#2176 In AI-Hypercomputer/maxtext;

· bjmsong opened on Aug 14, 2025

Regression: HF JSONL training halts early with

bugSomething isn't workingSomething isn't working

Status: Open.

#2147 In AI-Hypercomputer/maxtext;

· peregilk opened on Aug 10, 2025