GPUs for ML research · Crypto payment without KYC
IteraGPU
Laboratory / IteraGPU support · Orders and configurations
IteraGPU / ML research

Let's document what's blocking you.

A configuration to clarify or a payment to review? Find your order, describe the steps you took and save the relevant information in your account.

Prepare a useful diagnostic.

Provide the experiment or case reference, the steps performed, the expected result and the observed result. Mask keys, private credentials and sensitive data.

Save the context in your account to keep the useful details with your case reference.

Save a private request

From symptom to a case you can examine.

A run exceeds memory

Note the phase that fails: loading, generation, backward pass or evaluation. Record the batch, the input lengths, the precision and the maximum value observed. Also keep a smaller case that works. The difference between the two helps distinguish the weight of the model from the needs related to the data.

Reviewing the memory budget

A measurement seems inconsistent

Include the framework and extension versions, the input dimensions and the timing method. State whether you included loading and transfers in the total time. A comparison becomes usable when you know exactly what was measured and what changed between two runs.

Rereading the measurement protocol

A payment is still pending

Find the order and the intent used: asset, network, amount and time of the instructions. If the transfer was made, report it with "I paid" on that intent. To document a difficulty, keep the order reference and describe the steps; your private keys and recovery phrases are never necessary.

Resuming payment tracking

The content of your runs remains under your control. Prefer a reduced example and redacted messages over a complete dataset. Your private request gathers the context you choose to communicate and can be found in your account.