PhalaCloud/GLM-5.2-W4AFP8 · Discussions

Resources

View closed (0)

8xH20 141GB cuda out of memory

#6 opened 7 days ago by

ERROR: Should have a `model_type` key in its config.json

#5 opened 7 days ago by

Any chances for A100?

#4 opened 9 days ago by

GLM-5.2-W4AFP8 on 8×H100: fp8_e4m3 KV cache produces corrupted output, while BF16 KV works correctly

#3 opened 9 days ago by

Is the quantize script opensource?

#2 opened 9 days ago by

KTransformers + SGLang

#1 opened 10 days ago by