# LLM Training Optimization on Qwen 3 1.7B

Source: [LLM Training Optimization on Qwen 3 1.7B](<https://www.sota2.com/research/sota/llm-training-optimization-on-qwen-3-1-7b>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T06:06:31.021Z

Evaluation of efficiency improvements (time and energy reduction) during the training process of the Qwen 3 1.7B model.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | LLM Training Optimization |
| Dataset (including any stated split) | Qwen 3 1.7B |
| Metric | Time Reduction |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 18 of 18 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Time Reduction](<https://www.sota2.com/research/sota/llm-training-optimization-on-qwen-3-1-7b.md?metric=Time%20Reduction>): Higher is better; unit: SCALAR.

- [Energy Reduction](<https://www.sota2.com/research/sota/llm-training-optimization-on-qwen-3-1-7b.md?metric=Energy%20Reduction>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Time Reduction | Result date | Report paper |
| --- | --- | --- | --- |
| Kareus | 0.149 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 0.148 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 0.122 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 0.105 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 0.094 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 0.09 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 0.08 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 0.056 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 0.039 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 0.017 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | 0.005 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | 0.001 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | 0 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | -0.001 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | -0.005 | Not recorded | 6978467ca2b28e761d87c477 |
| Megatron-LM + Perseus | -0.005 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | -0.005 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | -0.204 | Not recorded | 6978467ca2b28e761d87c477 |

### Kareus

Time Reduction: 0.149. Result ID: 69991afb5698665ec4818d2d.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Kareus

Time Reduction: 0.148. Result ID: 69991afb5698665ec4818d30.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Time Reduction: 0.122. Result ID: 69991afb5698665ec4818d2a.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Time Reduction: 0.105. Result ID: 69991afb5698665ec4818d39.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: 0.094. Result ID: 69991afb5698665ec4818d2f.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: 0.09. Result ID: 69991afb5698665ec4818d2c.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Kareus

Time Reduction: 0.08. Result ID: 69991afb5698665ec4818d36.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: 0.056. Result ID: 69991afb5698665ec4818d29.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: 0.039. Result ID: 69991afb5698665ec4818d38.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: 0.017. Result ID: 69991afb5698665ec4818d35.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: 0.005. Result ID: 69991afb5698665ec4818d34.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: 0.001. Result ID: 69991afb5698665ec4818d2b.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"8K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: 0. Result ID: 69991afb5698665ec4818d37.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: -0.001. Result ID: 69991afb5698665ec4818d2e.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"16","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: -0.005. Result ID: 69991afb5698665ec4818d28.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"TP8","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Megatron-LM + Perseus

Time Reduction: -0.005. Result ID: 69991afb5698665ec4818d31.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Time Reduction: -0.005. Result ID: 69991afb5698665ec4818d33.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Time Reduction: -0.204. Result ID: 69991afb5698665ec4818d32.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism":"CP2TP4","µBatch Size":"8","Sequence Length":"4K"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.