|
|
@@ -1,4 +1,7 @@
|
|
|
-# GLM-5.3-Flash on 4× DGX Spark — aidendle94 sparkrun image (TP=4). DEFERRED.
|
|
|
+# GLM-5.3-Flash on 4× DGX Spark — aidendle94 sparkrun image (TP=4).
|
|
|
+# 2026-09-07: retargeted at production-1.2 (hub, parity with the 159 2x lane);
|
|
|
+# zai-org FP8 weights (306 G) now cached in the NAS hub — the old DEFERRED
|
|
|
+# blocker (pending 328 GB download) is gone. All 4 nodes free (pluto switch).
|
|
|
#
|
|
|
# aiden's production 4-node profile, verbatim translation of
|
|
|
# https://aidenle.com/recipes/glm-5-3-flash-4x-dgx-spark into our 4x launcher
|
|
|
@@ -23,8 +26,8 @@
|
|
|
# before exec; if our boot rejects it in the KV coordinator, add a mods script.
|
|
|
|
|
|
engine = "vllm"
|
|
|
-name = "GLM-5.3-Flash aiden image (vLLM, 4x Spark, DFlash2-7, FP8) [DEFERRED]"
|
|
|
-docker_image = "sparkrun-vllm-glm53-flash-aiden:glm53"
|
|
|
+name = "GLM-5.3-Flash aiden image (vLLM, 4x Spark, MTP-5, FP8)"
|
|
|
+docker_image = "aidendle94/sparkrun-vllm-glm53-gb10:production-1.2" # hub tag (parity with the benched 2x lane)
|
|
|
docker_image_dir = "glm53-flash-aiden"
|
|
|
model = "zai-org/GLM-5.3-Flash"
|
|
|
required_nodes = 4
|
