Shihan Qu

zenmagnets's activity

• All
• Models
• Datasets
• Spaces
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in lukealonso/Qwen3.5-397B-A17B-NVFP4 4 days ago

How to run on vLLM for 4xSM120

#1 opened 4 days ago by

zenmagnets

New activity in lukealonso/MiniMax-M2.5-NVFP4 17 days ago

Here's the vLLM recipe I'm using with 2x RTX Pro 6000

👍 3

17

#1 opened 29 days ago by

zenmagnets

New activity in Ex0bit/Kimi-K2.5-PRISM-REAP-530B-A32B 18 days ago

Anyone get this working on 4x RTX 6000 Pro?

👀 1

#1 opened 18 days ago by

zenmagnets

New activity in GadflyII/MiniMax-M2.1-NVFP4 23 days ago

Throughput NVFP4 on Dual 6000 Blackwells

#2 opened 23 days ago by

zenmagnets

New activity in vincentzed-hf/Qwen3.5-397B-A17B-NVFP4 25 days ago

Anyone try this on 4x RTX 6000 Pro yet?

52

#1 opened 26 days ago by

zenmagnets

New activity in mlx-community/Qwen3.5-397B-A17B-nvfp4 26 days ago

I wish it would fit in 2x6000 PRO!

1

#2 opened 27 days ago by

mtcl

New activity in lukealonso/MiniMax-M2.5-NVFP4 29 days ago

"w1_weight_scale_2 must match w3_weight_scale_2. Accuracy may be affected."

👍 1

20

#2 opened 29 days ago by

zenmagnets

New activity in GadflyII/GLM-4.7-Flash-NVFP4 about 2 months ago

Wasn't able to recreate MMLU-Pro benchmarks

5

#5 opened about 2 months ago by

zenmagnets

New activity in zai-org/GLM-4.7-Flash about 2 months ago

Enormous KV-cache size?

👍 ➕ 6

23

#3 opened about 2 months ago by

nephepritou

New activity in GadflyII/GLM-4.7-Flash-NVFP4 about 2 months ago

Really appreciate that you ran performance comparison tests with BF16!

3

#2 opened about 2 months ago by

zenmagnets

New activity in marksverdhei/GLM-4.7-Flash-FP8 about 2 months ago

Performance comps with BF16?

1

#3 opened about 2 months ago by

zenmagnets

New activity in cyankiwi/GLM-4.7-Flash-AWQ-4bit about 2 months ago

Any plans for a 6bit or 8bit version?

1

#3 opened about 2 months ago by

zenmagnets

New activity in marksverdhei/GLM-4.7-Flash-FP8 about 2 months ago

If 8bit, why shaped like 16 bit

2

#2 opened about 2 months ago by

zenmagnets

New activity in nvidia/Qwen3-30B-A3B-NVFP4 3 months ago

6 months since intro of NVFP4, and it's basically still a myth

1

#4 opened 3 months ago by

zenmagnets

New activity in RESMP-DEV/Qwen3-Next-80B-A3B-Instruct-NVFP4 3 months ago

Works with vllm? Any recommendations or howtos?

7

#1 opened 5 months ago by

DrRos

New activity in ig1/Qwen3-Next-80B-A3B-Instruct-NVFP4 3 months ago

Anyone get this to work?

#1 opened 3 months ago by

zenmagnets

New activity in Intel/Qwen3-Next-80B-A3B-Thinking-int4-AutoRound 4 months ago

What Token Generation are you guys getting with CPU only?

#3 opened 4 months ago by

zenmagnets