閸ャ垽妲﹂張鈧潻鎴濆枀鐎规碍濡窤I缂傛牜鐖滄禒锝囨倞闂嗗棙鍨氶崚鐧圛/CD濞翠焦鎸夌痪澶歌厬閿涘瞼娲伴惃鍕Ц鐠佲晛鎮楃粩顖涘絹娴溿倓鍞惍浣告倵閼奉亜濮╃挧棰佺闁秷顫夐懠鍐╊梾閺屻儱鎷伴崡鏇炲帗濞村鐦悽鐔稿灇閵嗗倹鍨滄潻娆掔珶闁鐎烽弮鍓佺€稉濠佺啊SkillAI鏉╂瑤閲滃鈧┃鎰般€嶉惄顔光偓鏂衡偓鏂跨暊閺€顖涘瘮婢舵碍膩閸ㄥ鎮楃粩顖ょ礉閺傚洦銆傞柌灞藉晸閻偓閳ユ粌绱戠粻鍗炲祮閻劉鈧縿鈧倷绲鹃惇鐔割劀娑撳﹥澧滈崥搴礉閹存垼绻涚紒顓′刊娴滃棔绗佹稉顏勬綑閿涘本濮岄懙鍙ョ啊閺佸瓨鏆ｆ稉銈呫亯閵?/p>
閻滎垰顣ㄩ敍姝厃thon 3.11閵嗕竼UDA 12.1閵嗕赋yTorch 2.1閿涘本婀伴崷鐧汿X 4090閵嗗倻鏁ら惃鍕侀崹瀣ЦHermes AI閻?B閹稿洣鎶ら悧鍫礄閸╄桨绨琇lama 3閿涘鈧倹瀵滈悶鍡氼嚛闁板秶鐤嗘稉宥囩暬娴ｅ函绱濇担鍡欘儑娑撯偓娑擃亪鏁婄拠顖氭皑閻╁瓨甯寸拋鈺傚灉閹冲吀绨￠妴?/p>
閹存垶瀵滈悡顪檏illAI閻ㄥ嫬鐣奸弬瑙勫瘹閸楁鍎寸純鍙夋箛閸旓紕顏敍灞芥儙閸斻劌鎮楃拫鍐暏 娣囶喖顦?03閸氬函绱濋崘宥堫嚞濮瑰倻娲块幒銉﹀Г 閺堫剙婀撮幒銊ф倞缂佸牅绨懗鍊熺獓娴滃棴绱濇担鍡涒偓鐔峰閸欘亝婀?5 tokens/s閿涘瞼鏁撴禍褏骞嗘晶鍐╃壌閺堫剚鐥呭▔鏇犳暏閵嗗倹鍨滈崘鍐茬暰閸掑洤鍩屾禍鎴狀伂閹恒劎鎮婇敍宀勨偓澶嬪DeepSeek API閳ユ柡鈧柨鐣犳笟鍨杹娑撴梹鏁幐涓爀rmes濡€崇€烽妴鍌氬讲SkillAI閻ㄥ嫭膩閸ㄥ鍘ょ純顕€鍣峰▽鈩冩箒DeepSeek妫板嫯顔曢敍灞惧灉閹靛濮╅弨閫涚啊 閸氬孩娼电紙绫梜illAI閻ㄥ嚘ssues閿涘苯褰傞悳鏉跨暊姒涙顓婚崣顏勭磻閺€鍙ョ啊OpenAI閵嗕竸nthropic缁涘鍤戞稉鐙縭ovider閵嗗倽顩﹂幒銉ュ弳DeepSeek瀵版鍤滃鍗炲晸娑撯偓娑?code>provider閹绘帊娆㈤敍灞惧灗閼板懐鏁?strong>Token娑擃叀娴?/strong>閺堝搫鍩楅垾鏂衡偓鏃堚偓姘崇箖娑撯偓娑擃亣浜ら柌蹇庡敩閻炲棙濡窪eepSeek閻ㄥ嫬婀撮崸鈧弰鐘茬殸閹存€琾enAI閺嶇厧绱￠妴鍌涘灉閼哄彉绨￠崡濠傜毈閺冭泛鍟撴禍鍡曢嚋Nginx閸欏秳鍞崝鐘辩瑐API Key闁繋绱堕敍宀€绮撴禍搴㈠灇閸旂喕鐨熼柅姘モ偓?/p> 閹舵鍚樼€瑰矁绻栨稉澶婂彠閿涘本鍨滅紒鍫滅艾鐠烘垿鈧矮绨￠弫缈犻嚋闁炬崘鐭鹃敍?/p> 閻滄澘婀崶銏ゆЕ濮ｅ繐銇夐懗鍊熷殰閸斻劎鏁撻幋鎰殶閻х偓娼崡鏇炲帗濞村鐦敍灞藉櫙绾喚宸奸崷?0%瀹革箑褰搁妴鍌濇閻掓湹绗夋俊渚碢T-4閿涘奔绲鹃幋鎰拱閸欘亝婀侀崡浣稿瀻娑斿绔撮妴?/p> 鏉╂瑦顐奸煪鈺佹綑鐠佲晜鍨滃ǎ鍗炲煝娴ｆ挷绱伴崚甯窗瀵偓濠ф€塈缂傛牜鐖滃銉ュ徔閸愬秮鈧粌绱戠粻鍗炲祮閻劉鈧繐绱濋惇鐔割劀閽€钘夋勾閺冩湹绡冩导姘礈娑撹櫣澧楅張顑锯偓浣规▔鐎涙ǜ鈧礁鎮楃粩顖氬悑鐎硅鈧嗏偓灞藉幢婢圭偨鈧倸顩ч弸婊€缍樻稊鐔锋躬閹碱厼缂撶猾璁虫妧閻ㄥ嫮閮寸紒鐕傜礉娑撳秴螛閸忓牏鏁炬總?strong>濡€崇€?閳?閹恒劎鎮婇崥搴ｎ伂 閳?Token娑擃叀娴?/strong>閻ㄥ嫯鐭剧痪鍨禈閵嗗倹娓堕崥搴㈠腹閼芥劒绔存稉顏嗐仦閸栨椽銆嶉惄?strong>skillaiPOST /v1/chat/completions閺冭绱濇潻鏂挎礀娴?code>Error 403: Rate limit exceeded閵嗗倹鍨滈弰搴㈡閻㈠疇顕禍鍜穒tHub Token閿涘本鈧簼绠炴导姘崇Т闂勬劧绱甸幒鎺撶叀娴滃棔绔寸亸蹇旀閿涘本澧犻崣鎴犲箛SkillAI姒涙顓绘导姘矤GitHub鐎圭偞妞傞幏澶婂絿閺堚偓閺傜増膩閸ㄥ娼堥柌宥冣偓鍌濃偓灞惧灉閺堫剙婀碒ermes AI閻ㄥ嫭膩閸ㄥ鏋冩禒鎯扮熅瀵板嫰鍘ょ純顔煎晸闁挎瑤绨￠敍灞筋嚤閼锋潙鐣犳稉鈧惄鎾櫢鐠囨洑绗呮潪鏂ょ礉閸戠姷顫楃亸杈╂暏閸忓绨￠崠鍨倳闂勬劙顤傞妴?/p>
鐟欙絽鍠?/strong>閿涙碍澧滈崝銊ョ殺濡€崇€烽弶鍐櫢閺€鎯у煂閹稿洤鐣鹃惄顔肩秿閿涘苯鑻熺拋鍓х枂OPENCLAW_NO_FETCH_MODEL=true閻滎垰顣ㄩ崣姗€鍣洪妴?br>
缁楊兛绨╅崸鎴窗CUDA OOM閿涘牊妯夌€涙ɑ瀛╅崙鐚寸礆
CUDA OOM: RuntimeError: CUDA out of memory. Tried to allocate 2.54 GiB閵嗗倹鍨滈弰鎯у幢24GB閺勬儳鐡ㄩ敍灞惧瘻閻炲棜顕╃捄?B濡€崇€?4K娑撳﹣绗呴弬鍥ㄧ梾闂傤噣顣介妴鍌涘笓娴滃棗宕愭径鈺嬬礉閸欐垹骞嘢killAI姒涙顓婚悽鈺32閸旂姾娴囧Ο鈥崇€烽敍宀冣偓瀛抏rmes AI閻ㄥ嫭膩閸ㄥ婀伴煬顐︽付鐟曚礁宕熼崡?6GB娴犮儰绗傞敍瀹杙32閻╁瓨甯撮悥鍡曠啊閵?/p>
鐟欙絽鍠?/strong>閿涙艾鎯庨崝銊ュ棘閺佹澘濮?code>--load-in-4bit閿涘本妯夌€涙ê宕伴悽銊╂閸?GB閵嗗倸褰熸径鏍ㄥΩPyTorch 2.1閻?code>torch.compile閸忚櫕甯€閳ユ柡鈧柨鐣犻崷銊﹀灉鏉╂瑥褰撮張鍝勬珤娑撳﹣绗夌粙鍐茬暰閿涘苯浼撶亸鏂垮毉閻?code>ConnectionTimeout鐎佃壈鍤UDA kernel缂傛牞鐦ф径杈Е閵?br>
缁楊兛绗侀崸鎴窗Token娑擃叀娴嗘稉宥epSeek API閻ㄥ嫬鍚嬬€?/h3>
provider鐎涙顔岄敍宀€绮ㄩ弸婊嗙殶閻劍妞傞幎?code>ConnectionTimeout: timed out after 30s閵?/p>
鐟欙絽鍠呴弬瑙勵攳閿涙碍鏆ｉ崥鍫濇倵閻ㄥ嫭娓舵担鍐茬杽鐠?/h2>
缂佹挸鐔敍姘紣閸忛攱妲稿锕€娅掗敍灞肩稻闁板秹鈧倹澧犻弰顖氬彠闁?/h2>