# LLaMA Factory — User Demand Report

**Week:** 2026-W15
**Generated:** 2026-04-06
**Issues analyzed:** 43 (41 included)
**Need clusters:** 1

## Top 10 User Needs

| Rank | Need | Issues | Score | Category | Examples |
| --- | --- | --- | --- | --- | --- |
| 1 | Distributed Training Infrastructure and Model Compatibility Fixes | 41 | 7.2 | Performance | [#10355](https://github.com/hiyouga/LlamaFactory/issues/10355), [#10351](https://github.com/hiyouga/LlamaFactory/issues/10351), [#10337](https://github.com/hiyouga/LlamaFactory/issues/10337) |

## Rising Needs

| Need | Rising Score | This Week | Category |
| --- | --- | --- | --- |
| Distributed Training Infrastructure and Model Compatibility Fixes | 1.3x | 41 | Performance |

## Category Breakdown

- **Performance**: 1 clusters

## All Need Clusters

### 1. Distributed Training Infrastructure and Model Compatibility Fixes

Users are experiencing multiple training infrastructure issues when fine-tuning large language and vision-language models, including distributed training errors (FSDP2, Ray, NCCL), data processing bugs, and model-specific compatibility problems with Qwen3.5 and Gemma4. These issues prevent reliable training across multi-GPU and specialized hardware (Apple Silicon MPS, Ascend NPUs) setups, requiring fixes to resource allocation, parameter offloading, and model loading to ensure stable fine-tuning workflows.

- **Volume:** 41 issues (23 open, 18 closed)
- **Demand Score:** 7.2
- **Avg Reactions:** 0 | **Avg Comments:** 1.2
- **Example issues:** [#10355](https://github.com/hiyouga/LlamaFactory/issues/10355), [#10351](https://github.com/hiyouga/LlamaFactory/issues/10351), [#10337](https://github.com/hiyouga/LlamaFactory/issues/10337), [#10317](https://github.com/hiyouga/LlamaFactory/issues/10317), [#10314](https://github.com/hiyouga/LlamaFactory/issues/10314)

---

*This report analyzes public GitHub issues only. It represents a signal from public issue discussions, not the full user base.*

*Generated by [ReadYourUsers](https://github.com/study8677/ReadYourUsers)*