🦁 Lion (LFMN)

Self Taught Financial Markets Expert | Open Source AI Enthusiast Developer | Camper Van Conversions | Remote Viewing | Love Food | Silence | I Build Stuff

[🦁 Lion (LFMN) @LionFMN]

Updated V1.4 new NAS Model Cassette Switching Dashboard with metrics out now. #NAS #Qwen #TensorFold #DGX #Spark #DGXSpark #AI #openSource

https://t.co/84Go9lqxks

[🦁 Lion (LFMN) @LionFMN]

@ashxhart @u1tra_instinct Sweet

[🦁 Lion (LFMN) @LionFMN]

New full AI Media Suite for Single DGX Spark, includes: Video with sound · images · music · voices · a director that writes and reviews every shot. Everything runs locally on one Spark.

https://t.co/aMwm5tSo9D

[🦁 Lion (LFMN) @LionFMN]

@NeoAIForecast Wow, sad times

[🦁 Lion (LFMN) @LionFMN]

@0xalfons You play too much.

[🦁 Lion (LFMN) @LionFMN]

@NeoAIForecast https://t.co/faMJftxz40

[🦁 Lion (LFMN) @LionFMN]

@NeoAIForecast Ridiculous

[🦁 Lion (LFMN) @LionFMN]

@NeoAIForecast 😂😂😂

[🦁 Lion (LFMN) @LionFMN]

@quimedesu 😆😆😆

[🦁 Lion (LFMN) @LionFMN]

@Hikari_07_jp Sounds like sparks

[🦁 Lion (LFMN) @LionFMN]

@RayFernando1337 @volatilemarkts Looks neat, I would run Linux on it.

[🦁 Lion (LFMN) @LionFMN]

@MiaAI_lab @TechMDAI Sweet setup!! Can we can some abliterared flash TensorFold EXL3 models ?

[🦁 Lion (LFMN) @LionFMN]

@MiaAI_lab Very nice!!

[🦁 Lion (LFMN) @LionFMN]

@nvidianewsroom @MiaAI_lab @Cisco Wow!!

[🦁 Lion (LFMN) @LionFMN]

@arikuschnir That’s really good

[🦁 Lion (LFMN) @LionFMN]

@Alibaba_Qwen - https://t.co/AYURH6VVhL 🔥🔥

[Qwen @Alibaba_Qwen]

Introducing QwenGyre, bringing elastic RL to extreme-long-horizon (XLong) agent training. ✨

It launches with two core designs: 🥳
- Elastic Scheduler: reallocates GPUs between rollout & training as demand shifts—live executions never interrupted. Up to 1.85× speedup over Colocate, 1.78× over Async.
- Trajectory Processor: turns branching executions (compaction, sub-agents, retries) into prefix-shared trajectory trees with bounded, deduplicated training samples. Partial progress after timeout still gets scored.

Scaled to Qwen3.8-2.4T-A95B with ~700K tokens per rollout, QwenGyre lifts NL2RepoBench from 52.5% to 58.5% in 48 steps, cutting end-to-end training time by ~44% vs Async.

🔗 Read the paper:
https://t.co/gbKPIiqosk

[🦁 Lion (LFMN) @LionFMN]

@MiaAI_lab Thanks

[🦁 Lion (LFMN) @LionFMN]

@jayleaton Cool, also check out my studio repo.

https://t.co/aMwm5tSo9D

[🦁 Lion (LFMN) @LionFMN]

@JensenHuang @satyanadella RTX Spark, should’ve been a Linux laptop with 256gb of ram, GB10 one PFLOP with QSFP port so that you can connect to DGX Spark to pool the ram.

[🦁 Lion (LFMN) @LionFMN]

@MiaAI_lab Scalable Vector Graphics (SVG) is good for flow charts and diagrams, not quite images.