Upload 4 files

adapter_config.json

@@ -0,0 +1,37 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": null,
+  "bias": "none",
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "q_proj",
+    "up_proj",
+    "down_proj",
+    "v_proj",
+    "gate_proj",
+    "o_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d467957619ee4d2bdb83638019594a159617024532a4a1e8cf5e075cb854b3d6
+size 113645520

stats.json

@@ -0,0 +1 @@
+{"world_size": 1, "epochs": 5, "steps": 12311, "seqs": 37730, "tokens": 319231170, "last_epoch_steps": 0, "last_epoch_seqs": 0, "last_epoch_tokens": 0, "wandb_run_id": null, "loss_ema": 0.3002565205329163, "loss_sum": 14.63201254606247, "mtp_loss_ema": 0, "mtp_loss_sum": 0, "eval_losses_avg": [0.5610001087188721, 0.5048913955688477, 0.47053754329681396, 0.45190441608428955, 0.4418374001979828, 0.43249592185020447, 0.42545926570892334, 0.41864991188049316, 0.41425052285194397, 0.40817946195602417, 0.40374594926834106, 0.40075385570526123, 0.3980681300163269, 0.39508408308029175, 0.39221280813217163, 0.3907524347305298, 0.3921872675418854, 0.3837128281593323, 0.3794099688529968, 0.376427561044693, 0.3762855529785156, 0.37402695417404175, 0.37452757358551025, 0.3725188970565796, 0.37079793214797974, 0.3700369894504547, 0.3709615468978882, 0.3706568479537964, 0.3697822690010071, 0.3684927225112915, 0.3663642406463623, 0.36859387159347534, 0.36443260312080383, 0.36567607522010803, 0.36369234323501587, 0.36343011260032654, 0.36226820945739746, 0.36137855052948, 0.35991984605789185, 0.35920682549476624, 0.3605673909187317, 0.35789066553115845, 0.3572970926761627, 0.3565283417701721, 0.35593554377555847, 0.3551166355609894, 0.35343286395072937, 0.35222485661506653, 0.35226884484291077, 0.35373204946517944, 0.3521794080734253, 0.34997570514678955, 0.34918665885925293, 0.3454052805900574, 0.34261175990104675, 0.3434743881225586, 0.34416264295578003, 0.3434031009674072, 0.3422328233718872, 0.3406336307525635, 0.34199005365371704, 0.34020373225212097, 0.3415184020996094, 0.3417931795120239, 0.3394811451435089, 0.3375788927078247, 0.33887428045272827, 0.33885619044303894, 0.33785247802734375, 0.33810728788375854, 0.33673322200775146, 0.3372512459754944, 0.3362613320350647, 0.33695682883262634, 0.3368929624557495, 0.33636438846588135, 0.3355070948600769, 0.3357078731060028, 0.335764616727829, 0.33407866954803467, 0.33588922023773193, 0.33655744791030884, 0.3369653522968292, 0.3340364098548889, 0.3246963322162628, 0.3209230899810791, 0.3206609785556793, 0.32160913944244385, 0.32312074303627014, 0.32204514741897583, 0.32188844680786133, 0.3163021206855774, 0.31087157130241394, 0.31016114354133606, 0.311825692653656, 0.3105613589286804, 0.31243187189102173, 0.31175440549850464, 0.3120635151863098, 0.31160879135131836, 0.3126956522464752, 0.3112768530845642, 0.3108038306236267, 0.31179696321487427, 0.31101173162460327, 0.31132179498672485, 0.3123365640640259, 0.31056535243988037, 0.31150004267692566, 0.31423747539520264, 0.3118616044521332, 0.31274425983428955, 0.3117019534111023, 0.31201428174972534, 0.310320109128952, 0.30921539664268494, 0.3118956387042999, 0.3119204342365265, 0.31232190132141113, 0.31061989068984985, 0.3100600242614746, 0.31054407358169556, 0.30912262201309204, 0.30989861488342285, 0.31028270721435547, 0.3095592260360718, 0.3089027404785156, 0.307538777589798, 0.30746710300445557, 0.3093253970146179, 0.3096756935119629, 0.3085050582885742, 0.30868273973464966, 0.31041547656059265, 0.3095286786556244, 0.31176236271858215, 0.3111018240451813, 0.31216779351234436, 0.31029725074768066, 0.30953705310821533, 0.30995041131973267, 0.30948156118392944, 0.3082911968231201, 0.30789703130722046, 0.30837932229042053, 0.3036888837814331, 0.30449342727661133, 0.3044801950454712, 0.3037782609462738, 0.30327272415161133, 0.30230772495269775, 0.30266323685646057, 0.30177220702171326, 0.3044753670692444, 0.30277615785598755, 0.30229926109313965, 0.28907638788223267, 0.29000407457351685, 0.29060858488082886, 0.2915119528770447, 0.29100745916366577, 0.2927902340888977, 0.29199475049972534, 0.291667103767395, 0.29148074984550476, 0.2914115786552429, 0.29156672954559326, 0.29354968667030334, 0.2931486964225769, 0.2931053638458252, 0.2927256226539612, 0.29235661029815674, 0.2942744195461273, 0.28014618158340454, 0.2816532850265503, 0.28118789196014404, 0.2816013693809509, 0.2829948663711548, 0.2806859612464905, 0.28284838795661926, 0.28310155868530273, 0.2848968207836151, 0.28369855880737305, 0.28331226110458374, 0.28365230560302734, 0.2839381992816925, 0.284939706325531, 0.28523918986320496, 0.28487709164619446, 0.2851211428642273, 0.2856754660606384, 0.28446775674819946, 0.28464919328689575, 0.2855614125728607, 0.28454339504241943, 0.2837054431438446, 0.2846434712409973, 0.28414827585220337, 0.27717214822769165, 0.2760953903198242, 0.27712297439575195, 0.2777382731437683, 0.2780223488807678, 0.27904653549194336, 0.27827349305152893, 0.2790948152542114, 0.27574795484542847, 0.2757904529571533, 0.2757342457771301, 0.27439844608306885, 0.27571040391921997, 0.27533799409866333, 0.2769715487957001, 0.27843236923217773, 0.2766357958316803, 0.27638906240463257, 0.27893999218940735, 0.27842608094215393, 0.2788024842739105, 0.27723610401153564, 0.2782502770423889, 0.27873528003692627, 0.27767884731292725, 0.2790883183479309, 0.2796929180622101, 0.27927255630493164, 0.27866899967193604, 0.27821335196495056, 0.27932673692703247, 0.279228538274765, 0.28107643127441406, 0.2797306776046753, 0.2779906094074249, 0.2776145935058594, 0.2783990502357483, 0.2774381935596466, 0.27822667360305786, 0.2787995934486389, 0.2786276340484619, 0.27987584471702576, 0.27905502915382385, 0.2793412208557129, 0.27926209568977356, 0.27896422147750854, 0.2794894278049469, 0.2793077826499939, 0.2783088684082031, 0.27888309955596924, 0.2791934013366699, 0.2793278694152832, 0.27968642115592957, 0.27820318937301636, 0.28029337525367737, 0.27859950065612793, 0.2795780599117279, 0.28048115968704224, 0.280427485704422, 0.2807706594467163, 0.2788315713405609, 0.27773764729499817, 0.2803802192211151, 0.27940282225608826, 0.2785906195640564, 0.27900430560112, 0.2787180542945862, 0.27773913741111755, 0.2768027186393738, 0.2789071202278137, 0.27807286381721497, 0.27699801325798035, 0.2767258882522583, 0.2774057984352112, 0.2775447964668274, 0.27816587686538696, 0.27724191546440125, 0.2768952548503876, 0.2762337923049927, 0.2778029441833496, 0.27749937772750854, 0.2790095806121826, 0.27951931953430176, 0.2790842056274414, 0.27886462211608887, 0.27977293729782104, 0.2800751328468323, 0.27981603145599365, 0.2801624834537506, 0.27548152208328247, 0.2698569893836975, 0.27024221420288086, 0.27042412757873535, 0.27283111214637756, 0.2705479562282562, 0.27076607942581177, 0.2709004282951355, 0.26551389694213867, 0.26364797353744507, 0.263397753238678, 0.2640073895454407, 0.265380859375, 0.26571333408355713, 0.26565560698509216, 0.2665339410305023, 0.2661752700805664, 0.26637908816337585, 0.2610814571380615, 0.260540634393692, 0.2605856657028198, 0.2616044878959656, 0.26215070486068726, 0.26231515407562256, 0.2636238634586334, 0.2628554701805115, 0.2614746689796448, 0.26335373520851135, 0.26251786947250366, 0.2634921669960022, 0.26226532459259033, 0.2629821300506592, 0.2629879415035248, 0.262317955493927, 0.26295220851898193, 0.26377803087234497, 0.26351088285446167, 0.26444000005722046, 0.2646160125732422, 0.2655675709247589, 0.2659316956996918, 0.25668612122535706, 0.2559153437614441, 0.2552800178527832, 0.25587373971939087, 0.2563169002532959, 0.25831276178359985, 0.25859174132347107, 0.25834396481513977, 0.2585926055908203, 0.25780755281448364, 0.25701069831848145, 0.2587772011756897, 0.25926315784454346, 0.2574707269668579, 0.25783592462539673, 0.25660941004753113, 0.2571391463279724, 0.2580687999725342, 0.2590261399745941, 0.26008340716362, 0.25994282960891724, 0.2580362558364868, 0.25853121280670166, 0.2595936954021454, 0.2603505849838257, 0.26063284277915955, 0.260660856962204, 0.26031625270843506, 0.2601197063922882, 0.26000845432281494, 0.26050418615341187, 0.2609716057777405, 0.2611197829246521, 0.2606065273284912, 0.260981947183609, 0.2612076997756958, 0.26037734746932983, 0.2591531276702881, 0.25922679901123047, 0.26191237568855286, 0.26040053367614746, 0.26206719875335693, 0.261478066444397, 0.26087579131126404, 0.2604089081287384, 0.2604910433292389, 0.2606698274612427, 0.2618263065814972, 0.262031614780426, 0.2630261778831482, 0.2622377276420593, 0.2609330415725708, 0.26188185811042786, 0.2603456676006317, 0.25973159074783325, 0.259585440158844, 0.26244455575942993, 0.26061832904815674, 0.26078975200653076, 0.26069992780685425, 0.2640811502933502, 0.2618805766105652, 0.26064300537109375, 0.2601834535598755, 0.26189398765563965, 0.2615225613117218, 0.261435866355896, 0.2611171007156372, 0.2521362602710724, 0.24892300367355347, 0.24912631511688232, 0.24876907467842102, 0.25021985173225403, 0.24988508224487305, 0.2512640058994293, 0.25163376331329346, 0.25194665789604187, 0.25178420543670654, 0.2520514726638794, 0.2524312138557434, 0.2535363435745239, 0.25253909826278687, 0.25506263971328735, 0.25346681475639343, 0.2455420345067978, 0.24027478694915771, 0.2418171763420105, 0.24285264313220978, 0.24226093292236328, 0.2425021529197693, 0.2346501350402832, 0.23435306549072266, 0.23393046855926514, 0.23489415645599365, 0.237639382481575, 0.23646040260791779, 0.23680032789707184, 0.23698526620864868, 0.23702797293663025, 0.23838567733764648, 0.23714913427829742, 0.23895326256752014, 0.23959609866142273, 0.23845794796943665, 0.23941966891288757, 0.2405790090560913, 0.23840680718421936, 0.24100440740585327, 0.24219948053359985, 0.2417522519826889, 0.24120807647705078, 0.23656053841114044, 0.23631799221038818, 0.23509681224822998, 0.23615342378616333, 0.2376002073287964, 0.2378394901752472, 0.23654112219810486, 0.23538830876350403, 0.23732344806194305, 0.23725390434265137, 0.23824244737625122, 0.23913386464118958, 0.239984393119812, 0.23957952857017517, 0.23879548907279968, 0.23969578742980957, 0.24030333757400513, 0.24028748273849487, 0.24151422083377838, 0.2402295172214508, 0.23924770951271057, 0.24008536338806152, 0.24215379357337952, 0.24175894260406494, 0.24376782774925232, 0.24264754354953766, 0.24140605330467224, 0.24315060675144196, 0.24173958599567413, 0.24184204638004303, 0.24234572052955627, 0.24295759201049805, 0.2436220794916153, 0.24337643384933472, 0.24258026480674744, 0.2437068074941635, 0.24370834231376648, 0.24192869663238525, 0.24394097924232483, 0.24497798085212708, 0.24383071064949036, 0.24423149228096008, 0.24454429745674133, 0.2449871450662613, 0.244979590177536, 0.2448999434709549, 0.2433655560016632, 0.24413061141967773, 0.24453336000442505, 0.244308203458786, 0.2465101182460785, 0.24689018726348877, 0.24597421288490295, 0.2464635670185089, 0.24726815521717072, 0.24706515669822693, 0.2445712685585022, 0.24532008171081543, 0.2451516091823578, 0.24564510583877563, 0.2446083128452301, 0.24728961288928986, 0.24634301662445068, 0.24637040495872498, 0.24707765877246857, 0.24755170941352844, 0.24776622653007507, 0.24673092365264893, 0.24729031324386597, 0.24709686636924744, 0.24775536358356476, 0.2482115626335144, 0.24802158772945404, 0.2482801079750061, 0.24875342845916748, 0.24861833453178406, 0.2485746443271637, 0.24964478611946106, 0.24708615243434906, 0.2474283128976822, 0.2486555278301239, 0.24843305349349976, 0.24741211533546448, 0.24818068742752075, 0.25031623244285583, 0.2503913640975952, 0.2500153183937073, 0.24956673383712769, 0.24837429821491241, 0.2508121728897095, 0.24827232956886292, 0.24882125854492188, 0.2501896321773529, 0.2500598132610321, 0.2479289174079895, 0.24831469357013702, 0.2473708987236023, 0.2479243278503418, 0.24900221824645996, 0.24855077266693115, 0.2499540150165558, 0.2509194016456604, 0.24889619648456573, 0.25032445788383484, 0.2517588138580322, 0.2509569525718689, 0.25229132175445557, 0.2512091398239136, 0.25054359436035156, 0.24954959750175476, 0.2505093216896057, 0.25099605321884155, 0.2508385181427002, 0.2513098418712616, 0.25176528096199036, 0.2522714138031006, 0.25095033645629883, 0.2513003945350647, 0.2509343922138214, 0.2497474104166031, 0.24607129395008087, 0.24415473639965057, 0.2447740137577057, 0.24494417011737823, 0.24485251307487488, 0.24551451206207275, 0.23374126851558685, 0.22283935546875, 0.22265005111694336, 0.22416949272155762, 0.22404591739177704, 0.22525861859321594, 0.227393239736557, 0.22756336629390717, 0.22830358147621155, 0.22803011536598206, 0.22936952114105225, 0.228801429271698, 0.2297038584947586, 0.22375184297561646, 0.22197255492210388, 0.22224310040473938, 0.22283637523651123, 0.22270287573337555, 0.2242068350315094, 0.22522249817848206, 0.2198946177959442, 0.21415162086486816, 0.21462097764015198, 0.21606507897377014, 0.21753254532814026, 0.21630291640758514, 0.2182672619819641, 0.21701419353485107, 0.21935546398162842, 0.21936553716659546, 0.2190672606229782, 0.21970009803771973, 0.22006285190582275, 0.2188853770494461, 0.21966463327407837, 0.2211584597826004, 0.2203599065542221, 0.22041162848472595, 0.22111916542053223, 0.2205551266670227, 0.22045589983463287, 0.22285713255405426, 0.2223195731639862, 0.22116735577583313, 0.22154134511947632, 0.2219294011592865, 0.2232166975736618, 0.22450965642929077, 0.2234843522310257, 0.22383981943130493, 0.2236488163471222, 0.22335860133171082, 0.22429300844669342]}

train_config.json

@@ -0,0 +1 @@
+{"base_model_dir": "/llm-downloader-destination/base/fireworks/gemma-3-27b-it/hf", "output_model_dir": "gs://fireworks-artifacts-mattrhorn27-f729ee-ab71a7/sftj-qbqzap5u/8cd991/gemmathinker/checkpoint", "checkpoint_dir": "/dev/shm/checkpoints", "gcs_checkpoint_dir": "gs://fireworks-fine-tuning-checkpoints/sftj-mattrhorn27-f729ee-qbqzap5u/checkpoints", "max_checkpoints_to_keep": 1, "checkpoint_interval": 3600, "train": true, "learning_rate": 0.0001, "learning_rate_warmup_steps": 0, "grad_accum_steps": 1, "epochs": 5, "early_stop": false, "seed": 42, "dataset_dir": "/mnt/staging/dataset", "eval_auto_carveout": true, "eval_dataset_dir": null, "train_limit": null, "max_context_len": 32000, "batch_size": 32768, "min_evals_per_epoch": 4, "precision": null, "status_file": "gs://fireworks-fine-tuning-job-status/sftj-mattrhorn27-f729ee-qbqzap5u", "billing_file": "gs://fireworks-fine-tuning-metadata/sftj-mattrhorn27-f729ee-qbqzap5u/billing", "wandb": false, "wandb_entity": null, "wandb_api_key": null, "wandb_project": null, "wandb_run_id": null, "profile": null, "weight_sharding": null, "activation_sharding": null, "empty_weights": false, "peft_addon_dir": null, "lora_rank": 8, "template_kind": "conversation", "template": null, "eval_train_ratio": 0.02, "mtp_config": {"enable_mtp": false, "freeze_base_model": false, "num_draft_tokens": 1}}