| import os,json,time,re,sys,concurrent.futures as cf |
| from openai import OpenAI |
| from items import ITEMS |
| c=OpenAI(base_url="https://router.huggingface.co/v1",api_key=os.environ.get("HF_TOKEN","x"),timeout=180) |
| SUFFIX={"en":"\n\nReply with only the final answer (a number or a single word), nothing else.", |
| "ta":"\n\nஇறுதி விடையை மட்டும் (ஒரு எண் அல்லது ஒரு சொல்) கூறவும், வேறு எதுவும் வேண்டாம்.", |
| "tl":"\n\nFinal answer mattum sollunga (oru number illa oru word), vera edhuvum venaam."} |
| N=int(sys.argv[1]) if len(sys.argv)>1 else 2 |
| jobs=[(it,l,s) for it in ITEMS for l in("en","ta","tl") for s in range(N)] |
| def run(j): |
| it,l,s=j |
| for a in range(8): |
| try: |
| t=time.time() |
| r=c.chat.completions.create(model="reflection/Beam-1.0-501B-A23B",messages=[{"role":"user","content":it[l]+SUFFIX[l]}],max_tokens=6000) |
| m=r.choices[0].message; u=r.usage |
| rd=u.completion_tokens_details.reasoning_tokens if u.completion_tokens_details else None |
| return dict(id=it["id"],cat=it["cat"],lang=l,sample=s,answer=it["answer"],aliases=it.get("aliases",[]),output=m.content,reasoning=getattr(m,"reasoning_content",None), |
| prompt_tokens=u.prompt_tokens,completion_tokens=u.completion_tokens,reasoning_tokens=rd,latency=round(time.time()-t,2),finish=r.choices[0].finish_reason) |
| except Exception as e: |
| err=str(e); time.sleep(5*(a+1)) |
| return dict(id=it["id"],lang=l,sample=s,error=err) |
| import os.path |
| prev={} |
| if os.path.exists("raw_results.json"): |
| for r in json.load(open("raw_results.json")): |
| if "error" not in r: prev[(r["id"],r["lang"],r["sample"])]=r |
| todo=[j for j in jobs if (j[0]["id"],j[1],j[2]) not in prev] |
| print("todo",len(todo),flush=True) |
| with cf.ThreadPoolExecutor(4) as ex: |
| for r in ex.map(run,todo): prev[(r["id"],r["lang"],r["sample"])]=r |
| out=list(prev.values()) |
| json.dump(out,open("raw_results.json","w"),ensure_ascii=False,indent=1) |
| print(len(out),sum('error' in r for r in out)) |
|
|
