Datasets:
File size: 1,861 Bytes
886afe0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | import json
req = {m["id"]: m for m in json.load(open("/mnt/localssd/seqroute1m/scripts/models_requested.json"))}
NOTHINK = {"enable_thinking": False}
MM = ["--limit-mm-per-prompt", json.dumps({"image": 0, "video": 0, "audio": 0})]
over = {
"qwen3_0p6b": dict(ctk=NOTHINK), "qwen3_1p7b": dict(ctk=NOTHINK), "qwen3_4b": dict(ctk=NOTHINK), "qwen3_8b": dict(ctk=NOTHINK), "qwen3_14b": dict(ctk=NOTHINK),
"qwen3_32b": dict(ctk=NOTHINK, tp=2),
"r1_distill_qwen_7b": dict(temperature=0.6, top_p=0.95), "r1_distill_qwen_14b": dict(temperature=0.6, top_p=0.95),
"phi4_mini_reasoning": dict(temperature=0.8, top_p=0.95),
"gpt_oss_20b": dict(temperature=1.0, ctk={"reasoning_effort": "low"}, serve_args=["--reasoning-parser", "openai_gptoss"]),
"olmo3_7b": dict(serve_args=["--chat-template", "/mnt/localssd/seqroute1m/templates/olmo3_7b.jinja"]),
"gemma3_4b": dict(serve_args=MM), "gemma3_12b": dict(serve_args=MM), "ministral3_8b": dict(serve_args=MM),
"magistral_small_24b": dict(temperature=0.7, top_p=0.95, serve_args=["--tokenizer-mode", "mistral", "--reasoning-parser", "mistral"] + MM, system=open("/mnt/localssd/models/magistral_small_24b/SYSTEM_PROMPT.txt").read()),
}
out = []
for mid, m in req.items():
o = over.get(mid, {})
out.append(dict(id=mid, family=m["family"], repo=m["repo"], requested=m["requested"], replacement_reason=m.get("reason"), params_b=m["params_b"], active_params_b=m.get("active_params_b", m["params_b"]),
reasoning=m["reasoning"], tp=o.get("tp", 1), temperature=o.get("temperature", 0.0), top_p=o.get("top_p", 1.0), chat_template_kwargs=o.get("ctk", {}),
serve_args=o.get("serve_args", []), max_model_len=16384, system=o.get("system"), fallback=m.get("fallback")))
json.dump(out, open("/mnt/localssd/seqroute1m/scripts/models_spec.json", "w"), indent=1)
print(len(out), "models")
|