I’m trying to use GLM‑5.3‑Flash with tool calling in WASTE 0.7.2, but the container appears to be configured only for K3-style XTML. As a result tool definitions are rejected.
Environment
WASTE version: 0.7.2
Container: v0
Backend: NEON, crc32 armv8, arm64
Model: glm5.3-flash — glm5-next, 45 layers, 288 experts, ctx 128000
Quant: experts VQ3R, trunk Q4G/Q8G/F32
OS: macOS (ARM)
error log excerpt:
Error: 400: {"message":"this container is served from its chat.json, which cannot express tool definitions because its tokenizer does not carry the Kimi K2 native tool markers","type":"invalid_request_error","param":"tools","code":null}
Chat config:
{
"prelude": "[gMASK]<sop>",
"system": ["<|system|>", ""],
"user": ["<|user|>", ""],
"assistant": ["<|assistant|>", ""],
"open": "<|assistant|>",
"think": ["<think>", "</think>"],
"stop": "<|user|>",
"image": "<|begin_of_image|><|image|><|end_of_image|>"
}
I’m trying to use GLM‑5.3‑Flash with tool calling in WASTE 0.7.2, but the container appears to be configured only for K3-style XTML. As a result tool definitions are rejected.
Environment
WASTE version: 0.7.2
Container: v0
Backend: NEON, crc32 armv8, arm64
Model: glm5.3-flash — glm5-next, 45 layers, 288 experts, ctx 128000
Quant: experts VQ3R, trunk Q4G/Q8G/F32
OS: macOS (ARM)
error log excerpt:
Error: 400: {"message":"this container is served from its chat.json, which cannot express tool definitions because its tokenizer does not carry the Kimi K2 native tool markers","type":"invalid_request_error","param":"tools","code":null}
Chat config: