RyanYr commited on
Commit
b3d5a21
1 Parent(s): 5735345

Training in progress, step 600, checkpoint

Browse files
last-checkpoint/global_step600/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d69cc7f02253096f45396aba167c1988e7728847825e9647246d6e08d61c2c20
3
+ size 7843036668
last-checkpoint/global_step600/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5455c8f3264a3ffe769e90e5192fd3463bba95e430f0f5ab93c3235fc65e44d8
3
+ size 7843043580
last-checkpoint/global_step600/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e3bdacca3246a6abbebb6c3dcd504e489b9113fb11af609e9d6682ac8574449
3
+ size 7843043004
last-checkpoint/global_step600/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e7d571105921123d324ab0f67ea9e7aff70d9c1248301354365beaf2cdf438d
3
+ size 7843043388
last-checkpoint/global_step600/mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61cc5d4071d6e42118cccee8b7660cf8f1da2d8e447c84806cbf5954ca8e9c8d
3
+ size 5228775200
last-checkpoint/latest CHANGED
@@ -1 +1 @@
1
- global_step300
 
1
+ global_step600
last-checkpoint/model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:66e5440ee43d7d965bd5ebefeb460324cd179609e49ba0224b4e9eba366671d0
3
  size 4988030368
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a4a62a2ca079e7006535ae0dfc6a8b327692d35ba7e5893eb61f1141b6f8a33
3
  size 4988030368
last-checkpoint/model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dc374d36b5ada303e3a3d9e8c2e7688d8c2494d9af7ea33766e5906178b09867
3
  size 1420344488
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c7cff20a673af4cb83e37e0780fb062da1d7d7d6735f5897e06aa1be3ad7d87
3
  size 1420344488
last-checkpoint/scheduler.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3c3ae9d50e8b8bedb2ad2de7843c691ae194ba67c96766e8f3cbe28cd2a287e6
3
  size 1064
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6febd8865f31f951fcb5c0712c5f0e814a0841d7eb33087fc8f49545c9d06e6
3
  size 1064
last-checkpoint/trainer_state.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
- "epoch": 0.3732794151955829,
5
  "eval_steps": 999999,
6
- "global_step": 300,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
@@ -2257,6 +2257,2256 @@
2257
  "rewards/margins": 1.7159814834594727,
2258
  "rewards/rejected": -1.9105104207992554,
2259
  "step": 300
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2260
  }
2261
  ],
2262
  "logging_steps": 2,
 
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
+ "epoch": 0.7465588303911658,
5
  "eval_steps": 999999,
6
+ "global_step": 600,
7
  "is_hyper_param_search": false,
8
  "is_local_process_zero": true,
9
  "is_world_process_zero": true,
 
2257
  "rewards/margins": 1.7159814834594727,
2258
  "rewards/rejected": -1.9105104207992554,
2259
  "step": 300
2260
+ },
2261
+ {
2262
+ "epoch": 0.37576794463022006,
2263
+ "grad_norm": 32.12793731689453,
2264
+ "learning_rate": 1.9125244284173493e-07,
2265
+ "logits/chosen": -9.93020248413086,
2266
+ "logits/rejected": -9.93803882598877,
2267
+ "logps/chosen": -16.381568908691406,
2268
+ "logps/rejected": -40.40930938720703,
2269
+ "loss": 0.4462,
2270
+ "rewards/accuracies": 0.65625,
2271
+ "rewards/chosen": 0.3086108863353729,
2272
+ "rewards/margins": 2.283296823501587,
2273
+ "rewards/rejected": -1.9746859073638916,
2274
+ "step": 302
2275
+ },
2276
+ {
2277
+ "epoch": 0.3782564740648573,
2278
+ "grad_norm": 16.97914695739746,
2279
+ "learning_rate": 1.9108100091877784e-07,
2280
+ "logits/chosen": -9.917818069458008,
2281
+ "logits/rejected": -9.916275978088379,
2282
+ "logps/chosen": -19.378997802734375,
2283
+ "logps/rejected": -38.81635284423828,
2284
+ "loss": 0.4085,
2285
+ "rewards/accuracies": 0.71875,
2286
+ "rewards/chosen": 0.04692254215478897,
2287
+ "rewards/margins": 1.8474431037902832,
2288
+ "rewards/rejected": -1.800520420074463,
2289
+ "step": 304
2290
+ },
2291
+ {
2292
+ "epoch": 0.3807450034994945,
2293
+ "grad_norm": 25.137126922607422,
2294
+ "learning_rate": 1.9090797360274307e-07,
2295
+ "logits/chosen": -9.911478996276855,
2296
+ "logits/rejected": -9.909990310668945,
2297
+ "logps/chosen": -12.876457214355469,
2298
+ "logps/rejected": -26.464027404785156,
2299
+ "loss": 0.4771,
2300
+ "rewards/accuracies": 0.6875,
2301
+ "rewards/chosen": 0.6640072464942932,
2302
+ "rewards/margins": 1.26017427444458,
2303
+ "rewards/rejected": -0.5961670279502869,
2304
+ "step": 306
2305
+ },
2306
+ {
2307
+ "epoch": 0.38323353293413176,
2308
+ "grad_norm": 34.36003494262695,
2309
+ "learning_rate": 1.907333639054147e-07,
2310
+ "logits/chosen": -9.968680381774902,
2311
+ "logits/rejected": -9.966692924499512,
2312
+ "logps/chosen": -16.23088836669922,
2313
+ "logps/rejected": -33.45315933227539,
2314
+ "loss": 0.4321,
2315
+ "rewards/accuracies": 0.65625,
2316
+ "rewards/chosen": 0.3378537893295288,
2317
+ "rewards/margins": 1.546026587486267,
2318
+ "rewards/rejected": -1.2081727981567383,
2319
+ "step": 308
2320
+ },
2321
+ {
2322
+ "epoch": 0.38572206236876894,
2323
+ "grad_norm": 28.8165225982666,
2324
+ "learning_rate": 1.9055717486612038e-07,
2325
+ "logits/chosen": -9.893492698669434,
2326
+ "logits/rejected": -9.893930435180664,
2327
+ "logps/chosen": -16.902034759521484,
2328
+ "logps/rejected": -25.581153869628906,
2329
+ "loss": 0.4179,
2330
+ "rewards/accuracies": 0.40625,
2331
+ "rewards/chosen": 0.30282124876976013,
2332
+ "rewards/margins": 0.8624454736709595,
2333
+ "rewards/rejected": -0.5596241354942322,
2334
+ "step": 310
2335
+ },
2336
+ {
2337
+ "epoch": 0.38821059180340617,
2338
+ "grad_norm": 59.199771881103516,
2339
+ "learning_rate": 1.9037940955167845e-07,
2340
+ "logits/chosen": -9.95876693725586,
2341
+ "logits/rejected": -9.965310096740723,
2342
+ "logps/chosen": -16.019397735595703,
2343
+ "logps/rejected": -35.83144760131836,
2344
+ "loss": 0.3918,
2345
+ "rewards/accuracies": 0.59375,
2346
+ "rewards/chosen": 0.3925018310546875,
2347
+ "rewards/margins": 1.9055030345916748,
2348
+ "rewards/rejected": -1.5130012035369873,
2349
+ "step": 312
2350
+ },
2351
+ {
2352
+ "epoch": 0.3906991212380434,
2353
+ "grad_norm": 33.99691390991211,
2354
+ "learning_rate": 1.9020007105634451e-07,
2355
+ "logits/chosen": -9.89972972869873,
2356
+ "logits/rejected": -9.901481628417969,
2357
+ "logps/chosen": -10.806205749511719,
2358
+ "logps/rejected": -24.527223587036133,
2359
+ "loss": 0.4215,
2360
+ "rewards/accuracies": 0.5625,
2361
+ "rewards/chosen": 0.8549016118049622,
2362
+ "rewards/margins": 1.0639026165008545,
2363
+ "rewards/rejected": -0.20900091528892517,
2364
+ "step": 314
2365
+ },
2366
+ {
2367
+ "epoch": 0.39318765067268063,
2368
+ "grad_norm": 34.28474807739258,
2369
+ "learning_rate": 1.9001916250175763e-07,
2370
+ "logits/chosen": -9.992077827453613,
2371
+ "logits/rejected": -9.988539695739746,
2372
+ "logps/chosen": -14.24827766418457,
2373
+ "logps/rejected": -37.29861831665039,
2374
+ "loss": 0.3997,
2375
+ "rewards/accuracies": 0.71875,
2376
+ "rewards/chosen": 0.5082123875617981,
2377
+ "rewards/margins": 2.078599452972412,
2378
+ "rewards/rejected": -1.5703872442245483,
2379
+ "step": 316
2380
+ },
2381
+ {
2382
+ "epoch": 0.3956761801073178,
2383
+ "grad_norm": 28.926748275756836,
2384
+ "learning_rate": 1.8983668703688597e-07,
2385
+ "logits/chosen": -9.971891403198242,
2386
+ "logits/rejected": -9.965632438659668,
2387
+ "logps/chosen": -15.308151245117188,
2388
+ "logps/rejected": -27.223262786865234,
2389
+ "loss": 0.4169,
2390
+ "rewards/accuracies": 0.53125,
2391
+ "rewards/chosen": 0.48706379532814026,
2392
+ "rewards/margins": 1.066077709197998,
2393
+ "rewards/rejected": -0.5790138840675354,
2394
+ "step": 318
2395
+ },
2396
+ {
2397
+ "epoch": 0.39816470954195504,
2398
+ "grad_norm": 90.87648010253906,
2399
+ "learning_rate": 1.896526478379719e-07,
2400
+ "logits/chosen": -9.971967697143555,
2401
+ "logits/rejected": -9.976310729980469,
2402
+ "logps/chosen": -17.316204071044922,
2403
+ "logps/rejected": -35.32228088378906,
2404
+ "loss": 0.5034,
2405
+ "rewards/accuracies": 0.46875,
2406
+ "rewards/chosen": 0.2696201503276825,
2407
+ "rewards/margins": 1.7619811296463013,
2408
+ "rewards/rejected": -1.4923608303070068,
2409
+ "step": 320
2410
+ },
2411
+ {
2412
+ "epoch": 0.4006532389765923,
2413
+ "grad_norm": 19.174610137939453,
2414
+ "learning_rate": 1.8946704810847686e-07,
2415
+ "logits/chosen": -10.004172325134277,
2416
+ "logits/rejected": -10.002047538757324,
2417
+ "logps/chosen": -14.942950248718262,
2418
+ "logps/rejected": -39.59999084472656,
2419
+ "loss": 0.4226,
2420
+ "rewards/accuracies": 0.71875,
2421
+ "rewards/chosen": 0.47852498292922974,
2422
+ "rewards/margins": 2.2441487312316895,
2423
+ "rewards/rejected": -1.765623688697815,
2424
+ "step": 322
2425
+ },
2426
+ {
2427
+ "epoch": 0.4031417684112295,
2428
+ "grad_norm": 17.031225204467773,
2429
+ "learning_rate": 1.8927989107902551e-07,
2430
+ "logits/chosen": -9.95960807800293,
2431
+ "logits/rejected": -9.958642959594727,
2432
+ "logps/chosen": -15.750935554504395,
2433
+ "logps/rejected": -33.05974578857422,
2434
+ "loss": 0.4574,
2435
+ "rewards/accuracies": 0.6875,
2436
+ "rewards/chosen": 0.35003215074539185,
2437
+ "rewards/margins": 1.4921305179595947,
2438
+ "rewards/rejected": -1.1420981884002686,
2439
+ "step": 324
2440
+ },
2441
+ {
2442
+ "epoch": 0.4056302978458667,
2443
+ "grad_norm": 6.252540111541748,
2444
+ "learning_rate": 1.890911800073495e-07,
2445
+ "logits/chosen": -10.024251937866211,
2446
+ "logits/rejected": -10.017958641052246,
2447
+ "logps/chosen": -16.81938934326172,
2448
+ "logps/rejected": -31.680240631103516,
2449
+ "loss": 0.4096,
2450
+ "rewards/accuracies": 0.5625,
2451
+ "rewards/chosen": 0.2728284001350403,
2452
+ "rewards/margins": 1.4476325511932373,
2453
+ "rewards/rejected": -1.1748039722442627,
2454
+ "step": 326
2455
+ },
2456
+ {
2457
+ "epoch": 0.4081188272805039,
2458
+ "grad_norm": 15.63929271697998,
2459
+ "learning_rate": 1.889009181782307e-07,
2460
+ "logits/chosen": -9.962908744812012,
2461
+ "logits/rejected": -9.967528343200684,
2462
+ "logps/chosen": -17.888456344604492,
2463
+ "logps/rejected": -43.57973861694336,
2464
+ "loss": 0.4092,
2465
+ "rewards/accuracies": 0.71875,
2466
+ "rewards/chosen": 0.19128906726837158,
2467
+ "rewards/margins": 2.440197467803955,
2468
+ "rewards/rejected": -2.248908519744873,
2469
+ "step": 328
2470
+ },
2471
+ {
2472
+ "epoch": 0.41060735671514115,
2473
+ "grad_norm": 30.135147094726562,
2474
+ "learning_rate": 1.8870910890344425e-07,
2475
+ "logits/chosen": -9.984109878540039,
2476
+ "logits/rejected": -9.985116004943848,
2477
+ "logps/chosen": -19.14948081970215,
2478
+ "logps/rejected": -38.880714416503906,
2479
+ "loss": 0.4442,
2480
+ "rewards/accuracies": 0.59375,
2481
+ "rewards/chosen": 0.038502052426338196,
2482
+ "rewards/margins": 1.8366103172302246,
2483
+ "rewards/rejected": -1.798108458518982,
2484
+ "step": 330
2485
+ },
2486
+ {
2487
+ "epoch": 0.4130958861497784,
2488
+ "grad_norm": 24.97335433959961,
2489
+ "learning_rate": 1.8851575552170062e-07,
2490
+ "logits/chosen": -9.994125366210938,
2491
+ "logits/rejected": -9.991096496582031,
2492
+ "logps/chosen": -13.679048538208008,
2493
+ "logps/rejected": -34.70538330078125,
2494
+ "loss": 0.3565,
2495
+ "rewards/accuracies": 0.8125,
2496
+ "rewards/chosen": 0.5989586114883423,
2497
+ "rewards/margins": 2.006279468536377,
2498
+ "rewards/rejected": -1.4073208570480347,
2499
+ "step": 332
2500
+ },
2501
+ {
2502
+ "epoch": 0.4155844155844156,
2503
+ "grad_norm": 30.876577377319336,
2504
+ "learning_rate": 1.8832086139858775e-07,
2505
+ "logits/chosen": -10.05797004699707,
2506
+ "logits/rejected": -10.054840087890625,
2507
+ "logps/chosen": -14.973274230957031,
2508
+ "logps/rejected": -37.49169158935547,
2509
+ "loss": 0.3823,
2510
+ "rewards/accuracies": 0.59375,
2511
+ "rewards/chosen": 0.49103009700775146,
2512
+ "rewards/margins": 2.1592507362365723,
2513
+ "rewards/rejected": -1.6682207584381104,
2514
+ "step": 334
2515
+ },
2516
+ {
2517
+ "epoch": 0.4180729450190528,
2518
+ "grad_norm": 21.57014274597168,
2519
+ "learning_rate": 1.881244299265122e-07,
2520
+ "logits/chosen": -10.034055709838867,
2521
+ "logits/rejected": -10.035269737243652,
2522
+ "logps/chosen": -17.542753219604492,
2523
+ "logps/rejected": -41.109371185302734,
2524
+ "loss": 0.359,
2525
+ "rewards/accuracies": 0.75,
2526
+ "rewards/chosen": 0.21824896335601807,
2527
+ "rewards/margins": 2.3213553428649902,
2528
+ "rewards/rejected": -2.1031060218811035,
2529
+ "step": 336
2530
+ },
2531
+ {
2532
+ "epoch": 0.42056147445369,
2533
+ "grad_norm": 41.96950912475586,
2534
+ "learning_rate": 1.8792646452464048e-07,
2535
+ "logits/chosen": -10.045391082763672,
2536
+ "logits/rejected": -10.043737411499023,
2537
+ "logps/chosen": -17.89931297302246,
2538
+ "logps/rejected": -28.092552185058594,
2539
+ "loss": 0.4804,
2540
+ "rewards/accuracies": 0.5,
2541
+ "rewards/chosen": 0.20682275295257568,
2542
+ "rewards/margins": 0.9021576046943665,
2543
+ "rewards/rejected": -0.695334792137146,
2544
+ "step": 338
2545
+ },
2546
+ {
2547
+ "epoch": 0.42305000388832725,
2548
+ "grad_norm": 41.85398864746094,
2549
+ "learning_rate": 1.8772696863883904e-07,
2550
+ "logits/chosen": -10.070752143859863,
2551
+ "logits/rejected": -10.066408157348633,
2552
+ "logps/chosen": -17.854961395263672,
2553
+ "logps/rejected": -29.23235321044922,
2554
+ "loss": 0.4977,
2555
+ "rewards/accuracies": 0.59375,
2556
+ "rewards/chosen": 0.18871577084064484,
2557
+ "rewards/margins": 0.9862973690032959,
2558
+ "rewards/rejected": -0.797581672668457,
2559
+ "step": 340
2560
+ },
2561
+ {
2562
+ "epoch": 0.4255385333229645,
2563
+ "grad_norm": 10.238739967346191,
2564
+ "learning_rate": 1.8752594574161477e-07,
2565
+ "logits/chosen": -10.056839942932129,
2566
+ "logits/rejected": -10.06256103515625,
2567
+ "logps/chosen": -16.154123306274414,
2568
+ "logps/rejected": -32.546260833740234,
2569
+ "loss": 0.4535,
2570
+ "rewards/accuracies": 0.59375,
2571
+ "rewards/chosen": 0.3366403877735138,
2572
+ "rewards/margins": 1.4845203161239624,
2573
+ "rewards/rejected": -1.1478800773620605,
2574
+ "step": 342
2575
+ },
2576
+ {
2577
+ "epoch": 0.42802706275760166,
2578
+ "grad_norm": 79.36576843261719,
2579
+ "learning_rate": 1.873233993320543e-07,
2580
+ "logits/chosen": -10.042964935302734,
2581
+ "logits/rejected": -10.046125411987305,
2582
+ "logps/chosen": -18.407886505126953,
2583
+ "logps/rejected": -34.48890686035156,
2584
+ "loss": 0.4725,
2585
+ "rewards/accuracies": 0.53125,
2586
+ "rewards/chosen": 0.17775145173072815,
2587
+ "rewards/margins": 1.4910780191421509,
2588
+ "rewards/rejected": -1.3133265972137451,
2589
+ "step": 344
2590
+ },
2591
+ {
2592
+ "epoch": 0.4305155921922389,
2593
+ "grad_norm": 11.346718788146973,
2594
+ "learning_rate": 1.8711933293576302e-07,
2595
+ "logits/chosen": -10.072295188903809,
2596
+ "logits/rejected": -10.066969871520996,
2597
+ "logps/chosen": -19.185441970825195,
2598
+ "logps/rejected": -39.912025451660156,
2599
+ "loss": 0.4647,
2600
+ "rewards/accuracies": 0.59375,
2601
+ "rewards/chosen": 0.06464995443820953,
2602
+ "rewards/margins": 1.9655919075012207,
2603
+ "rewards/rejected": -1.9009418487548828,
2604
+ "step": 346
2605
+ },
2606
+ {
2607
+ "epoch": 0.4330041216268761,
2608
+ "grad_norm": 26.568485260009766,
2609
+ "learning_rate": 1.8691375010480394e-07,
2610
+ "logits/chosen": -10.095144271850586,
2611
+ "logits/rejected": -10.09247875213623,
2612
+ "logps/chosen": -18.933631896972656,
2613
+ "logps/rejected": -34.512794494628906,
2614
+ "loss": 0.4134,
2615
+ "rewards/accuracies": 0.625,
2616
+ "rewards/chosen": 0.03263702988624573,
2617
+ "rewards/margins": 1.3736205101013184,
2618
+ "rewards/rejected": -1.340983510017395,
2619
+ "step": 348
2620
+ },
2621
+ {
2622
+ "epoch": 0.43549265106151336,
2623
+ "grad_norm": 29.14364242553711,
2624
+ "learning_rate": 1.8670665441763578e-07,
2625
+ "logits/chosen": -10.078664779663086,
2626
+ "logits/rejected": -10.079458236694336,
2627
+ "logps/chosen": -14.332991600036621,
2628
+ "logps/rejected": -27.881126403808594,
2629
+ "loss": 0.3661,
2630
+ "rewards/accuracies": 0.59375,
2631
+ "rewards/chosen": 0.5518330931663513,
2632
+ "rewards/margins": 1.2446177005767822,
2633
+ "rewards/rejected": -0.6927847862243652,
2634
+ "step": 350
2635
+ },
2636
+ {
2637
+ "epoch": 0.43798118049615054,
2638
+ "grad_norm": 83.77201843261719,
2639
+ "learning_rate": 1.8649804947905055e-07,
2640
+ "logits/chosen": -10.119363784790039,
2641
+ "logits/rejected": -10.119543075561523,
2642
+ "logps/chosen": -20.100664138793945,
2643
+ "logps/rejected": -34.80051803588867,
2644
+ "loss": 0.4812,
2645
+ "rewards/accuracies": 0.625,
2646
+ "rewards/chosen": -0.044760026037693024,
2647
+ "rewards/margins": 1.4141199588775635,
2648
+ "rewards/rejected": -1.4588799476623535,
2649
+ "step": 352
2650
+ },
2651
+ {
2652
+ "epoch": 0.44046970993078777,
2653
+ "grad_norm": 27.13106918334961,
2654
+ "learning_rate": 1.86287938920111e-07,
2655
+ "logits/chosen": -10.142522811889648,
2656
+ "logits/rejected": -10.139254570007324,
2657
+ "logps/chosen": -14.554240226745605,
2658
+ "logps/rejected": -32.03384780883789,
2659
+ "loss": 0.3947,
2660
+ "rewards/accuracies": 0.65625,
2661
+ "rewards/chosen": 0.5086873769760132,
2662
+ "rewards/margins": 1.5936012268066406,
2663
+ "rewards/rejected": -1.0849136114120483,
2664
+ "step": 354
2665
+ },
2666
+ {
2667
+ "epoch": 0.442958239365425,
2668
+ "grad_norm": 18.255369186401367,
2669
+ "learning_rate": 1.8607632639808722e-07,
2670
+ "logits/chosen": -10.121744155883789,
2671
+ "logits/rejected": -10.129932403564453,
2672
+ "logps/chosen": -15.290966033935547,
2673
+ "logps/rejected": -44.672088623046875,
2674
+ "loss": 0.3356,
2675
+ "rewards/accuracies": 0.65625,
2676
+ "rewards/chosen": 0.44472238421440125,
2677
+ "rewards/margins": 2.8495678901672363,
2678
+ "rewards/rejected": -2.404845714569092,
2679
+ "step": 356
2680
+ },
2681
+ {
2682
+ "epoch": 0.44544676880006223,
2683
+ "grad_norm": 30.86745262145996,
2684
+ "learning_rate": 1.8586321559639316e-07,
2685
+ "logits/chosen": -10.045269012451172,
2686
+ "logits/rejected": -10.052282333374023,
2687
+ "logps/chosen": -12.905274391174316,
2688
+ "logps/rejected": -24.89774513244629,
2689
+ "loss": 0.4592,
2690
+ "rewards/accuracies": 0.46875,
2691
+ "rewards/chosen": 0.6781768202781677,
2692
+ "rewards/margins": 1.1038376092910767,
2693
+ "rewards/rejected": -0.4256606698036194,
2694
+ "step": 358
2695
+ },
2696
+ {
2697
+ "epoch": 0.4479352982346994,
2698
+ "grad_norm": 30.82040786743164,
2699
+ "learning_rate": 1.856486102245224e-07,
2700
+ "logits/chosen": -10.125411033630371,
2701
+ "logits/rejected": -10.129536628723145,
2702
+ "logps/chosen": -15.23957347869873,
2703
+ "logps/rejected": -30.655988693237305,
2704
+ "loss": 0.4446,
2705
+ "rewards/accuracies": 0.59375,
2706
+ "rewards/chosen": 0.46949002146720886,
2707
+ "rewards/margins": 1.5255208015441895,
2708
+ "rewards/rejected": -1.0560307502746582,
2709
+ "step": 360
2710
+ },
2711
+ {
2712
+ "epoch": 0.45042382766933664,
2713
+ "grad_norm": 19.052501678466797,
2714
+ "learning_rate": 1.8543251401798373e-07,
2715
+ "logits/chosen": -10.178777694702148,
2716
+ "logits/rejected": -10.177990913391113,
2717
+ "logps/chosen": -14.576260566711426,
2718
+ "logps/rejected": -24.758251190185547,
2719
+ "loss": 0.4447,
2720
+ "rewards/accuracies": 0.5,
2721
+ "rewards/chosen": 0.5348184704780579,
2722
+ "rewards/margins": 1.0102537870407104,
2723
+ "rewards/rejected": -0.475435346364975,
2724
+ "step": 362
2725
+ },
2726
+ {
2727
+ "epoch": 0.4529123571039739,
2728
+ "grad_norm": 30.12545394897461,
2729
+ "learning_rate": 1.852149307382358e-07,
2730
+ "logits/chosen": -10.141602516174316,
2731
+ "logits/rejected": -10.141905784606934,
2732
+ "logps/chosen": -12.735381126403809,
2733
+ "logps/rejected": -26.4114990234375,
2734
+ "loss": 0.4658,
2735
+ "rewards/accuracies": 0.75,
2736
+ "rewards/chosen": 0.681415319442749,
2737
+ "rewards/margins": 1.30160653591156,
2738
+ "rewards/rejected": -0.6201913356781006,
2739
+ "step": 364
2740
+ },
2741
+ {
2742
+ "epoch": 0.4554008865386111,
2743
+ "grad_norm": 16.3223876953125,
2744
+ "learning_rate": 1.8499586417262207e-07,
2745
+ "logits/chosen": -10.104928970336914,
2746
+ "logits/rejected": -10.118807792663574,
2747
+ "logps/chosen": -13.549112319946289,
2748
+ "logps/rejected": -45.761314392089844,
2749
+ "loss": 0.4057,
2750
+ "rewards/accuracies": 0.65625,
2751
+ "rewards/chosen": 0.6719975471496582,
2752
+ "rewards/margins": 3.1581478118896484,
2753
+ "rewards/rejected": -2.4861507415771484,
2754
+ "step": 366
2755
+ },
2756
+ {
2757
+ "epoch": 0.4578894159732483,
2758
+ "grad_norm": 45.04033660888672,
2759
+ "learning_rate": 1.847753181343046e-07,
2760
+ "logits/chosen": -10.148914337158203,
2761
+ "logits/rejected": -10.151677131652832,
2762
+ "logps/chosen": -19.287796020507812,
2763
+ "logps/rejected": -31.361068725585938,
2764
+ "loss": 0.4228,
2765
+ "rewards/accuracies": 0.6875,
2766
+ "rewards/chosen": 0.07429736852645874,
2767
+ "rewards/margins": 1.160717487335205,
2768
+ "rewards/rejected": -1.0864201784133911,
2769
+ "step": 368
2770
+ },
2771
+ {
2772
+ "epoch": 0.4603779454078855,
2773
+ "grad_norm": 15.918371200561523,
2774
+ "learning_rate": 1.8455329646219765e-07,
2775
+ "logits/chosen": -10.12684440612793,
2776
+ "logits/rejected": -10.128087997436523,
2777
+ "logps/chosen": -15.174302101135254,
2778
+ "logps/rejected": -35.70402526855469,
2779
+ "loss": 0.3747,
2780
+ "rewards/accuracies": 0.65625,
2781
+ "rewards/chosen": 0.4587894082069397,
2782
+ "rewards/margins": 1.966828465461731,
2783
+ "rewards/rejected": -1.5080389976501465,
2784
+ "step": 370
2785
+ },
2786
+ {
2787
+ "epoch": 0.46286647484252275,
2788
+ "grad_norm": 25.0142822265625,
2789
+ "learning_rate": 1.8432980302090116e-07,
2790
+ "logits/chosen": -10.077094078063965,
2791
+ "logits/rejected": -10.081006050109863,
2792
+ "logps/chosen": -13.792490005493164,
2793
+ "logps/rejected": -29.12347412109375,
2794
+ "loss": 0.3935,
2795
+ "rewards/accuracies": 0.59375,
2796
+ "rewards/chosen": 0.641818642616272,
2797
+ "rewards/margins": 1.463584065437317,
2798
+ "rewards/rejected": -0.8217653632164001,
2799
+ "step": 372
2800
+ },
2801
+ {
2802
+ "epoch": 0.46535500427716,
2803
+ "grad_norm": 36.42277145385742,
2804
+ "learning_rate": 1.8410484170063315e-07,
2805
+ "logits/chosen": -10.140583992004395,
2806
+ "logits/rejected": -10.138748168945312,
2807
+ "logps/chosen": -18.85614585876465,
2808
+ "logps/rejected": -29.82183837890625,
2809
+ "loss": 0.4635,
2810
+ "rewards/accuracies": 0.71875,
2811
+ "rewards/chosen": 0.08829772472381592,
2812
+ "rewards/margins": 1.046219825744629,
2813
+ "rewards/rejected": -0.9579219818115234,
2814
+ "step": 374
2815
+ },
2816
+ {
2817
+ "epoch": 0.4678435337117972,
2818
+ "grad_norm": 17.340988159179688,
2819
+ "learning_rate": 1.8387841641716222e-07,
2820
+ "logits/chosen": -10.150492668151855,
2821
+ "logits/rejected": -10.134815216064453,
2822
+ "logps/chosen": -15.305471420288086,
2823
+ "logps/rejected": -35.03512191772461,
2824
+ "loss": 0.3603,
2825
+ "rewards/accuracies": 0.71875,
2826
+ "rewards/chosen": 0.46442386507987976,
2827
+ "rewards/margins": 1.800581693649292,
2828
+ "rewards/rejected": -1.3361577987670898,
2829
+ "step": 376
2830
+ },
2831
+ {
2832
+ "epoch": 0.4703320631464344,
2833
+ "grad_norm": 13.729105949401855,
2834
+ "learning_rate": 1.8365053111173924e-07,
2835
+ "logits/chosen": -10.110525131225586,
2836
+ "logits/rejected": -10.113253593444824,
2837
+ "logps/chosen": -18.30204200744629,
2838
+ "logps/rejected": -40.33465576171875,
2839
+ "loss": 0.4382,
2840
+ "rewards/accuracies": 0.625,
2841
+ "rewards/chosen": 0.1614200472831726,
2842
+ "rewards/margins": 1.9729737043380737,
2843
+ "rewards/rejected": -1.8115534782409668,
2844
+ "step": 378
2845
+ },
2846
+ {
2847
+ "epoch": 0.4728205925810716,
2848
+ "grad_norm": 63.69013214111328,
2849
+ "learning_rate": 1.8342118975102884e-07,
2850
+ "logits/chosen": -10.150627136230469,
2851
+ "logits/rejected": -10.151355743408203,
2852
+ "logps/chosen": -16.167449951171875,
2853
+ "logps/rejected": -43.09782791137695,
2854
+ "loss": 0.421,
2855
+ "rewards/accuracies": 0.71875,
2856
+ "rewards/chosen": 0.3487149178981781,
2857
+ "rewards/margins": 2.5346381664276123,
2858
+ "rewards/rejected": -2.1859230995178223,
2859
+ "step": 380
2860
+ },
2861
+ {
2862
+ "epoch": 0.47530912201570885,
2863
+ "grad_norm": 29.9747314453125,
2864
+ "learning_rate": 1.831903963270404e-07,
2865
+ "logits/chosen": -10.166023254394531,
2866
+ "logits/rejected": -10.171725273132324,
2867
+ "logps/chosen": -15.274872779846191,
2868
+ "logps/rejected": -39.947025299072266,
2869
+ "loss": 0.4205,
2870
+ "rewards/accuracies": 0.6875,
2871
+ "rewards/chosen": 0.45054560899734497,
2872
+ "rewards/margins": 2.247271776199341,
2873
+ "rewards/rejected": -1.796726107597351,
2874
+ "step": 382
2875
+ },
2876
+ {
2877
+ "epoch": 0.4777976514503461,
2878
+ "grad_norm": 22.047622680664062,
2879
+ "learning_rate": 1.829581548570584e-07,
2880
+ "logits/chosen": -10.139307022094727,
2881
+ "logits/rejected": -10.138298034667969,
2882
+ "logps/chosen": -13.432818412780762,
2883
+ "logps/rejected": -27.648792266845703,
2884
+ "loss": 0.3883,
2885
+ "rewards/accuracies": 0.46875,
2886
+ "rewards/chosen": 0.6289280652999878,
2887
+ "rewards/margins": 1.316452980041504,
2888
+ "rewards/rejected": -0.687524676322937,
2889
+ "step": 384
2890
+ },
2891
+ {
2892
+ "epoch": 0.48028618088498326,
2893
+ "grad_norm": 26.30389404296875,
2894
+ "learning_rate": 1.827244693835727e-07,
2895
+ "logits/chosen": -10.177475929260254,
2896
+ "logits/rejected": -10.185769081115723,
2897
+ "logps/chosen": -15.253499031066895,
2898
+ "logps/rejected": -37.98992919921875,
2899
+ "loss": 0.4129,
2900
+ "rewards/accuracies": 0.59375,
2901
+ "rewards/chosen": 0.38813167810440063,
2902
+ "rewards/margins": 2.0788075923919678,
2903
+ "rewards/rejected": -1.690676212310791,
2904
+ "step": 386
2905
+ },
2906
+ {
2907
+ "epoch": 0.4827747103196205,
2908
+ "grad_norm": 43.2744255065918,
2909
+ "learning_rate": 1.82489343974208e-07,
2910
+ "logits/chosen": -10.16571044921875,
2911
+ "logits/rejected": -10.16866683959961,
2912
+ "logps/chosen": -13.284004211425781,
2913
+ "logps/rejected": -41.31989669799805,
2914
+ "loss": 0.4216,
2915
+ "rewards/accuracies": 0.8125,
2916
+ "rewards/chosen": 0.6636133790016174,
2917
+ "rewards/margins": 2.6542181968688965,
2918
+ "rewards/rejected": -1.9906047582626343,
2919
+ "step": 388
2920
+ },
2921
+ {
2922
+ "epoch": 0.4852632397542577,
2923
+ "grad_norm": 31.072860717773438,
2924
+ "learning_rate": 1.8225278272165316e-07,
2925
+ "logits/chosen": -10.171157836914062,
2926
+ "logits/rejected": -10.169686317443848,
2927
+ "logps/chosen": -14.49439525604248,
2928
+ "logps/rejected": -31.08568572998047,
2929
+ "loss": 0.4284,
2930
+ "rewards/accuracies": 0.71875,
2931
+ "rewards/chosen": 0.5197519659996033,
2932
+ "rewards/margins": 1.5475728511810303,
2933
+ "rewards/rejected": -1.0278208255767822,
2934
+ "step": 390
2935
+ },
2936
+ {
2937
+ "epoch": 0.48775176918889496,
2938
+ "grad_norm": 18.515132904052734,
2939
+ "learning_rate": 1.8201478974358993e-07,
2940
+ "logits/chosen": -10.156515121459961,
2941
+ "logits/rejected": -10.155784606933594,
2942
+ "logps/chosen": -14.687192916870117,
2943
+ "logps/rejected": -39.48086929321289,
2944
+ "loss": 0.3517,
2945
+ "rewards/accuracies": 0.5625,
2946
+ "rewards/chosen": 0.5118663907051086,
2947
+ "rewards/margins": 2.3870129585266113,
2948
+ "rewards/rejected": -1.875146746635437,
2949
+ "step": 392
2950
+ },
2951
+ {
2952
+ "epoch": 0.49024029862353213,
2953
+ "grad_norm": 13.063323974609375,
2954
+ "learning_rate": 1.817753691826212e-07,
2955
+ "logits/chosen": -10.209580421447754,
2956
+ "logits/rejected": -10.209765434265137,
2957
+ "logps/chosen": -16.342390060424805,
2958
+ "logps/rejected": -39.23331832885742,
2959
+ "loss": 0.3614,
2960
+ "rewards/accuracies": 0.75,
2961
+ "rewards/chosen": 0.34056925773620605,
2962
+ "rewards/margins": 2.223206043243408,
2963
+ "rewards/rejected": -1.8826367855072021,
2964
+ "step": 394
2965
+ },
2966
+ {
2967
+ "epoch": 0.49272882805816937,
2968
+ "grad_norm": 12.84131908416748,
2969
+ "learning_rate": 1.8153452520619894e-07,
2970
+ "logits/chosen": -10.244853973388672,
2971
+ "logits/rejected": -10.24595832824707,
2972
+ "logps/chosen": -15.772315979003906,
2973
+ "logps/rejected": -40.866546630859375,
2974
+ "loss": 0.3374,
2975
+ "rewards/accuracies": 0.75,
2976
+ "rewards/chosen": 0.358955442905426,
2977
+ "rewards/margins": 2.356814384460449,
2978
+ "rewards/rejected": -1.997859001159668,
2979
+ "step": 396
2980
+ },
2981
+ {
2982
+ "epoch": 0.4952173574928066,
2983
+ "grad_norm": 15.765931129455566,
2984
+ "learning_rate": 1.8129226200655175e-07,
2985
+ "logits/chosen": -10.189569473266602,
2986
+ "logits/rejected": -10.18414306640625,
2987
+ "logps/chosen": -12.982626914978027,
2988
+ "logps/rejected": -35.74321746826172,
2989
+ "loss": 0.3984,
2990
+ "rewards/accuracies": 0.71875,
2991
+ "rewards/chosen": 0.6791236400604248,
2992
+ "rewards/margins": 2.1384410858154297,
2993
+ "rewards/rejected": -1.459317684173584,
2994
+ "step": 398
2995
+ },
2996
+ {
2997
+ "epoch": 0.49770588692744383,
2998
+ "grad_norm": 18.18596839904785,
2999
+ "learning_rate": 1.8104858380061177e-07,
3000
+ "logits/chosen": -10.160497665405273,
3001
+ "logits/rejected": -10.162215232849121,
3002
+ "logps/chosen": -15.011301040649414,
3003
+ "logps/rejected": -35.6241569519043,
3004
+ "loss": 0.3414,
3005
+ "rewards/accuracies": 0.75,
3006
+ "rewards/chosen": 0.4255256950855255,
3007
+ "rewards/margins": 1.9121336936950684,
3008
+ "rewards/rejected": -1.4866080284118652,
3009
+ "step": 400
3010
+ },
3011
+ {
3012
+ "epoch": 0.5001944163620811,
3013
+ "grad_norm": 32.82666015625,
3014
+ "learning_rate": 1.8080349482994128e-07,
3015
+ "logits/chosen": -10.219493865966797,
3016
+ "logits/rejected": -10.233976364135742,
3017
+ "logps/chosen": -11.769068717956543,
3018
+ "logps/rejected": -45.81578063964844,
3019
+ "loss": 0.4127,
3020
+ "rewards/accuracies": 0.625,
3021
+ "rewards/chosen": 0.8247968554496765,
3022
+ "rewards/margins": 3.2704405784606934,
3023
+ "rewards/rejected": -2.445643424987793,
3024
+ "step": 402
3025
+ },
3026
+ {
3027
+ "epoch": 0.5026829457967182,
3028
+ "grad_norm": 63.731163024902344,
3029
+ "learning_rate": 1.8055699936065897e-07,
3030
+ "logits/chosen": -10.156604766845703,
3031
+ "logits/rejected": -10.159815788269043,
3032
+ "logps/chosen": -13.73430061340332,
3033
+ "logps/rejected": -32.176734924316406,
3034
+ "loss": 0.3614,
3035
+ "rewards/accuracies": 0.65625,
3036
+ "rewards/chosen": 0.6367957592010498,
3037
+ "rewards/margins": 1.7830668687820435,
3038
+ "rewards/rejected": -1.1462712287902832,
3039
+ "step": 404
3040
+ },
3041
+ {
3042
+ "epoch": 0.5051714752313554,
3043
+ "grad_norm": 17.729053497314453,
3044
+ "learning_rate": 1.8030910168336556e-07,
3045
+ "logits/chosen": -10.223674774169922,
3046
+ "logits/rejected": -10.2188720703125,
3047
+ "logps/chosen": -12.967857360839844,
3048
+ "logps/rejected": -31.420896530151367,
3049
+ "loss": 0.3841,
3050
+ "rewards/accuracies": 0.625,
3051
+ "rewards/chosen": 0.7218824625015259,
3052
+ "rewards/margins": 1.7282061576843262,
3053
+ "rewards/rejected": -1.0063236951828003,
3054
+ "step": 406
3055
+ },
3056
+ {
3057
+ "epoch": 0.5076600046659927,
3058
+ "grad_norm": 20.5875301361084,
3059
+ "learning_rate": 1.8005980611306923e-07,
3060
+ "logits/chosen": -10.229300498962402,
3061
+ "logits/rejected": -10.224252700805664,
3062
+ "logps/chosen": -11.18129825592041,
3063
+ "logps/rejected": -34.46989822387695,
3064
+ "loss": 0.3851,
3065
+ "rewards/accuracies": 0.625,
3066
+ "rewards/chosen": 0.8643984198570251,
3067
+ "rewards/margins": 2.1738507747650146,
3068
+ "rewards/rejected": -1.3094525337219238,
3069
+ "step": 408
3070
+ },
3071
+ {
3072
+ "epoch": 0.5101485341006299,
3073
+ "grad_norm": 17.28754425048828,
3074
+ "learning_rate": 1.798091169891104e-07,
3075
+ "logits/chosen": -10.220544815063477,
3076
+ "logits/rejected": -10.224019050598145,
3077
+ "logps/chosen": -14.894316673278809,
3078
+ "logps/rejected": -32.1892204284668,
3079
+ "loss": 0.3394,
3080
+ "rewards/accuracies": 0.65625,
3081
+ "rewards/chosen": 0.5480198860168457,
3082
+ "rewards/margins": 1.6748688220977783,
3083
+ "rewards/rejected": -1.1268489360809326,
3084
+ "step": 410
3085
+ },
3086
+ {
3087
+ "epoch": 0.5126370635352672,
3088
+ "grad_norm": 8.287089347839355,
3089
+ "learning_rate": 1.7955703867508632e-07,
3090
+ "logits/chosen": -10.298615455627441,
3091
+ "logits/rejected": -10.291454315185547,
3092
+ "logps/chosen": -11.289877891540527,
3093
+ "logps/rejected": -32.0279426574707,
3094
+ "loss": 0.3391,
3095
+ "rewards/accuracies": 0.625,
3096
+ "rewards/chosen": 0.8290339112281799,
3097
+ "rewards/margins": 1.9380998611450195,
3098
+ "rewards/rejected": -1.1090660095214844,
3099
+ "step": 412
3100
+ },
3101
+ {
3102
+ "epoch": 0.5151255929699043,
3103
+ "grad_norm": 32.0596923828125,
3104
+ "learning_rate": 1.7930357555877498e-07,
3105
+ "logits/chosen": -10.227286338806152,
3106
+ "logits/rejected": -10.220754623413086,
3107
+ "logps/chosen": -11.198624610900879,
3108
+ "logps/rejected": -31.559532165527344,
3109
+ "loss": 0.4074,
3110
+ "rewards/accuracies": 0.5625,
3111
+ "rewards/chosen": 0.8517309427261353,
3112
+ "rewards/margins": 1.7727643251419067,
3113
+ "rewards/rejected": -0.9210332632064819,
3114
+ "step": 414
3115
+ },
3116
+ {
3117
+ "epoch": 0.5176141224045415,
3118
+ "grad_norm": 12.409061431884766,
3119
+ "learning_rate": 1.7904873205205884e-07,
3120
+ "logits/chosen": -10.299351692199707,
3121
+ "logits/rejected": -10.29948616027832,
3122
+ "logps/chosen": -15.70001220703125,
3123
+ "logps/rejected": -32.885704040527344,
3124
+ "loss": 0.3608,
3125
+ "rewards/accuracies": 0.625,
3126
+ "rewards/chosen": 0.3757689595222473,
3127
+ "rewards/margins": 1.5837979316711426,
3128
+ "rewards/rejected": -1.2080289125442505,
3129
+ "step": 416
3130
+ },
3131
+ {
3132
+ "epoch": 0.5201026518391788,
3133
+ "grad_norm": 60.90420913696289,
3134
+ "learning_rate": 1.7879251259084803e-07,
3135
+ "logits/chosen": -10.276573181152344,
3136
+ "logits/rejected": -10.284320831298828,
3137
+ "logps/chosen": -20.019948959350586,
3138
+ "logps/rejected": -55.10029602050781,
3139
+ "loss": 0.3512,
3140
+ "rewards/accuracies": 0.8125,
3141
+ "rewards/chosen": 0.028229378163814545,
3142
+ "rewards/margins": 3.422579765319824,
3143
+ "rewards/rejected": -3.394350528717041,
3144
+ "step": 418
3145
+ },
3146
+ {
3147
+ "epoch": 0.522591181273816,
3148
+ "grad_norm": 68.20502471923828,
3149
+ "learning_rate": 1.7853492163500303e-07,
3150
+ "logits/chosen": -10.290699005126953,
3151
+ "logits/rejected": -10.287884712219238,
3152
+ "logps/chosen": -19.97264289855957,
3153
+ "logps/rejected": -38.0428581237793,
3154
+ "loss": 0.3907,
3155
+ "rewards/accuracies": 0.59375,
3156
+ "rewards/chosen": -0.060215648263692856,
3157
+ "rewards/margins": 1.6576623916625977,
3158
+ "rewards/rejected": -1.7178781032562256,
3159
+ "step": 420
3160
+ },
3161
+ {
3162
+ "epoch": 0.5250797107084533,
3163
+ "grad_norm": 39.193138122558594,
3164
+ "learning_rate": 1.7827596366825717e-07,
3165
+ "logits/chosen": -10.288238525390625,
3166
+ "logits/rejected": -10.28443717956543,
3167
+ "logps/chosen": -19.868900299072266,
3168
+ "logps/rejected": -34.88875961303711,
3169
+ "loss": 0.4174,
3170
+ "rewards/accuracies": 0.65625,
3171
+ "rewards/chosen": 0.014708593487739563,
3172
+ "rewards/margins": 1.446516513824463,
3173
+ "rewards/rejected": -1.4318079948425293,
3174
+ "step": 422
3175
+ },
3176
+ {
3177
+ "epoch": 0.5275682401430905,
3178
+ "grad_norm": 129.97903442382812,
3179
+ "learning_rate": 1.780156431981385e-07,
3180
+ "logits/chosen": -10.253746032714844,
3181
+ "logits/rejected": -10.267844200134277,
3182
+ "logps/chosen": -19.551225662231445,
3183
+ "logps/rejected": -44.87137222290039,
3184
+ "loss": 0.3927,
3185
+ "rewards/accuracies": 0.5625,
3186
+ "rewards/chosen": 0.01799405738711357,
3187
+ "rewards/margins": 2.44464373588562,
3188
+ "rewards/rejected": -2.426649332046509,
3189
+ "step": 424
3190
+ },
3191
+ {
3192
+ "epoch": 0.5300567695777276,
3193
+ "grad_norm": 19.153640747070312,
3194
+ "learning_rate": 1.7775396475589143e-07,
3195
+ "logits/chosen": -10.24868392944336,
3196
+ "logits/rejected": -10.248597145080566,
3197
+ "logps/chosen": -22.83911895751953,
3198
+ "logps/rejected": -36.29539489746094,
3199
+ "loss": 0.4036,
3200
+ "rewards/accuracies": 0.625,
3201
+ "rewards/chosen": -0.32136890292167664,
3202
+ "rewards/margins": 1.164161205291748,
3203
+ "rewards/rejected": -1.485530138015747,
3204
+ "step": 426
3205
+ },
3206
+ {
3207
+ "epoch": 0.5325452990123649,
3208
+ "grad_norm": 58.614959716796875,
3209
+ "learning_rate": 1.7749093289639766e-07,
3210
+ "logits/chosen": -10.283756256103516,
3211
+ "logits/rejected": -10.286145210266113,
3212
+ "logps/chosen": -22.322837829589844,
3213
+ "logps/rejected": -46.103111267089844,
3214
+ "loss": 0.4529,
3215
+ "rewards/accuracies": 0.59375,
3216
+ "rewards/chosen": -0.2568221092224121,
3217
+ "rewards/margins": 2.2835302352905273,
3218
+ "rewards/rejected": -2.5403523445129395,
3219
+ "step": 428
3220
+ },
3221
+ {
3222
+ "epoch": 0.5350338284470021,
3223
+ "grad_norm": 28.71092987060547,
3224
+ "learning_rate": 1.7722655219809715e-07,
3225
+ "logits/chosen": -10.26553726196289,
3226
+ "logits/rejected": -10.262138366699219,
3227
+ "logps/chosen": -22.724733352661133,
3228
+ "logps/rejected": -45.04386520385742,
3229
+ "loss": 0.3481,
3230
+ "rewards/accuracies": 0.6875,
3231
+ "rewards/chosen": -0.2697991132736206,
3232
+ "rewards/margins": 2.1471457481384277,
3233
+ "rewards/rejected": -2.416944980621338,
3234
+ "step": 430
3235
+ },
3236
+ {
3237
+ "epoch": 0.5375223578816393,
3238
+ "grad_norm": 9.071741104125977,
3239
+ "learning_rate": 1.7696082726290823e-07,
3240
+ "logits/chosen": -10.279452323913574,
3241
+ "logits/rejected": -10.281999588012695,
3242
+ "logps/chosen": -24.338172912597656,
3243
+ "logps/rejected": -54.75845718383789,
3244
+ "loss": 0.3392,
3245
+ "rewards/accuracies": 0.78125,
3246
+ "rewards/chosen": -0.4605673551559448,
3247
+ "rewards/margins": 2.8818345069885254,
3248
+ "rewards/rejected": -3.3424017429351807,
3249
+ "step": 432
3250
+ },
3251
+ {
3252
+ "epoch": 0.5400108873162766,
3253
+ "grad_norm": 30.557416915893555,
3254
+ "learning_rate": 1.7669376271614753e-07,
3255
+ "logits/chosen": -10.22339153289795,
3256
+ "logits/rejected": -10.23956298828125,
3257
+ "logps/chosen": -18.716413497924805,
3258
+ "logps/rejected": -49.26506042480469,
3259
+ "loss": 0.3328,
3260
+ "rewards/accuracies": 0.75,
3261
+ "rewards/chosen": 0.14645807445049286,
3262
+ "rewards/margins": 2.979231595993042,
3263
+ "rewards/rejected": -2.832773208618164,
3264
+ "step": 434
3265
+ },
3266
+ {
3267
+ "epoch": 0.5424994167509137,
3268
+ "grad_norm": 26.827932357788086,
3269
+ "learning_rate": 1.7642536320644963e-07,
3270
+ "logits/chosen": -10.279085159301758,
3271
+ "logits/rejected": -10.276779174804688,
3272
+ "logps/chosen": -21.30095100402832,
3273
+ "logps/rejected": -47.970794677734375,
3274
+ "loss": 0.4278,
3275
+ "rewards/accuracies": 0.75,
3276
+ "rewards/chosen": -0.19874787330627441,
3277
+ "rewards/margins": 2.43815279006958,
3278
+ "rewards/rejected": -2.6369009017944336,
3279
+ "step": 436
3280
+ },
3281
+ {
3282
+ "epoch": 0.544987946185551,
3283
+ "grad_norm": 14.18700122833252,
3284
+ "learning_rate": 1.761556334056859e-07,
3285
+ "logits/chosen": -10.297018051147461,
3286
+ "logits/rejected": -10.291116714477539,
3287
+ "logps/chosen": -19.83533477783203,
3288
+ "logps/rejected": -33.00040817260742,
3289
+ "loss": 0.3998,
3290
+ "rewards/accuracies": 0.53125,
3291
+ "rewards/chosen": 0.010127291083335876,
3292
+ "rewards/margins": 1.244734525680542,
3293
+ "rewards/rejected": -1.2346073389053345,
3294
+ "step": 438
3295
+ },
3296
+ {
3297
+ "epoch": 0.5474764756201882,
3298
+ "grad_norm": 29.784212112426758,
3299
+ "learning_rate": 1.7588457800888338e-07,
3300
+ "logits/chosen": -10.262384414672852,
3301
+ "logits/rejected": -10.26101016998291,
3302
+ "logps/chosen": -22.732542037963867,
3303
+ "logps/rejected": -43.187095642089844,
3304
+ "loss": 0.4454,
3305
+ "rewards/accuracies": 0.65625,
3306
+ "rewards/chosen": -0.3034452199935913,
3307
+ "rewards/margins": 1.910001516342163,
3308
+ "rewards/rejected": -2.213446617126465,
3309
+ "step": 440
3310
+ },
3311
+ {
3312
+ "epoch": 0.5499650050548254,
3313
+ "grad_norm": 37.86104202270508,
3314
+ "learning_rate": 1.7561220173414295e-07,
3315
+ "logits/chosen": -10.271835327148438,
3316
+ "logits/rejected": -10.269705772399902,
3317
+ "logps/chosen": -21.843021392822266,
3318
+ "logps/rejected": -39.968994140625,
3319
+ "loss": 0.4206,
3320
+ "rewards/accuracies": 0.65625,
3321
+ "rewards/chosen": -0.23939497768878937,
3322
+ "rewards/margins": 1.706654667854309,
3323
+ "rewards/rejected": -1.9460498094558716,
3324
+ "step": 442
3325
+ },
3326
+ {
3327
+ "epoch": 0.5524535344894627,
3328
+ "grad_norm": 102.59848022460938,
3329
+ "learning_rate": 1.7533850932255716e-07,
3330
+ "logits/chosen": -10.290996551513672,
3331
+ "logits/rejected": -10.284740447998047,
3332
+ "logps/chosen": -24.096864700317383,
3333
+ "logps/rejected": -47.94099426269531,
3334
+ "loss": 0.3529,
3335
+ "rewards/accuracies": 0.625,
3336
+ "rewards/chosen": -0.40619534254074097,
3337
+ "rewards/margins": 2.2394251823425293,
3338
+ "rewards/rejected": -2.645620584487915,
3339
+ "step": 444
3340
+ },
3341
+ {
3342
+ "epoch": 0.5549420639240998,
3343
+ "grad_norm": 112.88279724121094,
3344
+ "learning_rate": 1.7506350553812788e-07,
3345
+ "logits/chosen": -10.27847671508789,
3346
+ "logits/rejected": -10.274407386779785,
3347
+ "logps/chosen": -21.205625534057617,
3348
+ "logps/rejected": -48.48828887939453,
3349
+ "loss": 0.3601,
3350
+ "rewards/accuracies": 0.71875,
3351
+ "rewards/chosen": -0.16596946120262146,
3352
+ "rewards/margins": 2.564211845397949,
3353
+ "rewards/rejected": -2.7301814556121826,
3354
+ "step": 446
3355
+ },
3356
+ {
3357
+ "epoch": 0.5574305933587371,
3358
+ "grad_norm": 17.369672775268555,
3359
+ "learning_rate": 1.747871951676832e-07,
3360
+ "logits/chosen": -10.271676063537598,
3361
+ "logits/rejected": -10.267562866210938,
3362
+ "logps/chosen": -19.17977523803711,
3363
+ "logps/rejected": -34.472694396972656,
3364
+ "loss": 0.4076,
3365
+ "rewards/accuracies": 0.5,
3366
+ "rewards/chosen": 0.009292155504226685,
3367
+ "rewards/margins": 1.3818503618240356,
3368
+ "rewards/rejected": -1.372558355331421,
3369
+ "step": 448
3370
+ },
3371
+ {
3372
+ "epoch": 0.5599191227933743,
3373
+ "grad_norm": 15.42812728881836,
3374
+ "learning_rate": 1.7450958302079425e-07,
3375
+ "logits/chosen": -10.265366554260254,
3376
+ "logits/rejected": -10.265597343444824,
3377
+ "logps/chosen": -21.165231704711914,
3378
+ "logps/rejected": -35.36635208129883,
3379
+ "loss": 0.4446,
3380
+ "rewards/accuracies": 0.625,
3381
+ "rewards/chosen": -0.1172541007399559,
3382
+ "rewards/margins": 1.3716140985488892,
3383
+ "rewards/rejected": -1.4888681173324585,
3384
+ "step": 450
3385
+ },
3386
+ {
3387
+ "epoch": 0.5624076522280115,
3388
+ "grad_norm": 10.552681922912598,
3389
+ "learning_rate": 1.7423067392969134e-07,
3390
+ "logits/chosen": -10.309708595275879,
3391
+ "logits/rejected": -10.30586051940918,
3392
+ "logps/chosen": -21.561840057373047,
3393
+ "logps/rejected": -42.782718658447266,
3394
+ "loss": 0.3014,
3395
+ "rewards/accuracies": 0.75,
3396
+ "rewards/chosen": -0.18897439539432526,
3397
+ "rewards/margins": 2.033334970474243,
3398
+ "rewards/rejected": -2.2223095893859863,
3399
+ "step": 452
3400
+ },
3401
+ {
3402
+ "epoch": 0.5648961816626488,
3403
+ "grad_norm": 34.31137466430664,
3404
+ "learning_rate": 1.7395047274917993e-07,
3405
+ "logits/chosen": -10.265932083129883,
3406
+ "logits/rejected": -10.26201057434082,
3407
+ "logps/chosen": -19.098529815673828,
3408
+ "logps/rejected": -37.488792419433594,
3409
+ "loss": 0.3513,
3410
+ "rewards/accuracies": 0.71875,
3411
+ "rewards/chosen": 0.060742493718862534,
3412
+ "rewards/margins": 1.7550294399261475,
3413
+ "rewards/rejected": -1.6942869424819946,
3414
+ "step": 454
3415
+ },
3416
+ {
3417
+ "epoch": 0.5673847110972859,
3418
+ "grad_norm": 33.42908477783203,
3419
+ "learning_rate": 1.736689843565562e-07,
3420
+ "logits/chosen": -10.311141967773438,
3421
+ "logits/rejected": -10.312514305114746,
3422
+ "logps/chosen": -17.21253204345703,
3423
+ "logps/rejected": -37.408626556396484,
3424
+ "loss": 0.3602,
3425
+ "rewards/accuracies": 0.625,
3426
+ "rewards/chosen": 0.28720271587371826,
3427
+ "rewards/margins": 1.9509572982788086,
3428
+ "rewards/rejected": -1.6637544631958008,
3429
+ "step": 456
3430
+ },
3431
+ {
3432
+ "epoch": 0.5698732405319231,
3433
+ "grad_norm": 22.34549903869629,
3434
+ "learning_rate": 1.7338621365152193e-07,
3435
+ "logits/chosen": -10.262404441833496,
3436
+ "logits/rejected": -10.266767501831055,
3437
+ "logps/chosen": -15.291158676147461,
3438
+ "logps/rejected": -48.40684509277344,
3439
+ "loss": 0.3249,
3440
+ "rewards/accuracies": 0.75,
3441
+ "rewards/chosen": 0.43968135118484497,
3442
+ "rewards/margins": 3.158674716949463,
3443
+ "rewards/rejected": -2.7189934253692627,
3444
+ "step": 458
3445
+ },
3446
+ {
3447
+ "epoch": 0.5723617699665604,
3448
+ "grad_norm": 28.7365665435791,
3449
+ "learning_rate": 1.731021655560995e-07,
3450
+ "logits/chosen": -10.315303802490234,
3451
+ "logits/rejected": -10.321941375732422,
3452
+ "logps/chosen": -11.67857837677002,
3453
+ "logps/rejected": -35.23655700683594,
3454
+ "loss": 0.3909,
3455
+ "rewards/accuracies": 0.625,
3456
+ "rewards/chosen": 0.7879737615585327,
3457
+ "rewards/margins": 2.1632347106933594,
3458
+ "rewards/rejected": -1.3752611875534058,
3459
+ "step": 460
3460
+ },
3461
+ {
3462
+ "epoch": 0.5748502994011976,
3463
+ "grad_norm": 112.85546112060547,
3464
+ "learning_rate": 1.728168450145459e-07,
3465
+ "logits/chosen": -10.332886695861816,
3466
+ "logits/rejected": -10.332941055297852,
3467
+ "logps/chosen": -10.49925422668457,
3468
+ "logps/rejected": -34.67118835449219,
3469
+ "loss": 0.4086,
3470
+ "rewards/accuracies": 0.71875,
3471
+ "rewards/chosen": 0.8904905319213867,
3472
+ "rewards/margins": 2.1655731201171875,
3473
+ "rewards/rejected": -1.2750823497772217,
3474
+ "step": 462
3475
+ },
3476
+ {
3477
+ "epoch": 0.5773388288358349,
3478
+ "grad_norm": 29.10820960998535,
3479
+ "learning_rate": 1.7253025699326704e-07,
3480
+ "logits/chosen": -10.357038497924805,
3481
+ "logits/rejected": -10.358717918395996,
3482
+ "logps/chosen": -8.701577186584473,
3483
+ "logps/rejected": -37.82337951660156,
3484
+ "loss": 0.374,
3485
+ "rewards/accuracies": 0.78125,
3486
+ "rewards/chosen": 1.13681161403656,
3487
+ "rewards/margins": 2.682598829269409,
3488
+ "rewards/rejected": -1.5457874536514282,
3489
+ "step": 464
3490
+ },
3491
+ {
3492
+ "epoch": 0.579827358270472,
3493
+ "grad_norm": 19.05025291442871,
3494
+ "learning_rate": 1.7224240648073093e-07,
3495
+ "logits/chosen": -10.339391708374023,
3496
+ "logits/rejected": -10.337575912475586,
3497
+ "logps/chosen": -11.895441055297852,
3498
+ "logps/rejected": -48.07970428466797,
3499
+ "loss": 0.4306,
3500
+ "rewards/accuracies": 0.8125,
3501
+ "rewards/chosen": 0.7691047787666321,
3502
+ "rewards/margins": 3.405965805053711,
3503
+ "rewards/rejected": -2.6368610858917236,
3504
+ "step": 466
3505
+ },
3506
+ {
3507
+ "epoch": 0.5823158877051092,
3508
+ "grad_norm": 26.512428283691406,
3509
+ "learning_rate": 1.719532984873811e-07,
3510
+ "logits/chosen": -10.381338119506836,
3511
+ "logits/rejected": -10.382423400878906,
3512
+ "logps/chosen": -11.953356742858887,
3513
+ "logps/rejected": -31.554622650146484,
3514
+ "loss": 0.3554,
3515
+ "rewards/accuracies": 0.53125,
3516
+ "rewards/chosen": 0.7555459141731262,
3517
+ "rewards/margins": 1.866694688796997,
3518
+ "rewards/rejected": -1.111148715019226,
3519
+ "step": 468
3520
+ },
3521
+ {
3522
+ "epoch": 0.5848044171397465,
3523
+ "grad_norm": 14.106731414794922,
3524
+ "learning_rate": 1.7166293804554926e-07,
3525
+ "logits/chosen": -10.35709285736084,
3526
+ "logits/rejected": -10.35936164855957,
3527
+ "logps/chosen": -11.882340431213379,
3528
+ "logps/rejected": -43.57575988769531,
3529
+ "loss": 0.3134,
3530
+ "rewards/accuracies": 0.78125,
3531
+ "rewards/chosen": 0.8108737468719482,
3532
+ "rewards/margins": 2.9975638389587402,
3533
+ "rewards/rejected": -2.186689853668213,
3534
+ "step": 470
3535
+ },
3536
+ {
3537
+ "epoch": 0.5872929465743837,
3538
+ "grad_norm": 10.351644515991211,
3539
+ "learning_rate": 1.7137133020936782e-07,
3540
+ "logits/chosen": -10.363170623779297,
3541
+ "logits/rejected": -10.355613708496094,
3542
+ "logps/chosen": -8.152997016906738,
3543
+ "logps/rejected": -31.83338165283203,
3544
+ "loss": 0.3955,
3545
+ "rewards/accuracies": 0.75,
3546
+ "rewards/chosen": 1.142722487449646,
3547
+ "rewards/margins": 2.2745003700256348,
3548
+ "rewards/rejected": -1.1317781209945679,
3549
+ "step": 472
3550
+ },
3551
+ {
3552
+ "epoch": 0.5897814760090209,
3553
+ "grad_norm": 43.42179870605469,
3554
+ "learning_rate": 1.7107848005468176e-07,
3555
+ "logits/chosen": -10.350369453430176,
3556
+ "logits/rejected": -10.345197677612305,
3557
+ "logps/chosen": -11.4368257522583,
3558
+ "logps/rejected": -32.16423416137695,
3559
+ "loss": 0.3702,
3560
+ "rewards/accuracies": 0.59375,
3561
+ "rewards/chosen": 0.8080503344535828,
3562
+ "rewards/margins": 1.9460852146148682,
3563
+ "rewards/rejected": -1.1380348205566406,
3564
+ "step": 474
3565
+ },
3566
+ {
3567
+ "epoch": 0.5922700054436582,
3568
+ "grad_norm": 19.114572525024414,
3569
+ "learning_rate": 1.7078439267896038e-07,
3570
+ "logits/chosen": -10.334250450134277,
3571
+ "logits/rejected": -10.345415115356445,
3572
+ "logps/chosen": -12.645822525024414,
3573
+ "logps/rejected": -44.80878829956055,
3574
+ "loss": 0.406,
3575
+ "rewards/accuracies": 0.78125,
3576
+ "rewards/chosen": 0.7018153667449951,
3577
+ "rewards/margins": 2.9746077060699463,
3578
+ "rewards/rejected": -2.272792339324951,
3579
+ "step": 476
3580
+ },
3581
+ {
3582
+ "epoch": 0.5947585348782953,
3583
+ "grad_norm": 52.6199951171875,
3584
+ "learning_rate": 1.7048907320120864e-07,
3585
+ "logits/chosen": -10.30965805053711,
3586
+ "logits/rejected": -10.303977012634277,
3587
+ "logps/chosen": -11.632840156555176,
3588
+ "logps/rejected": -35.78714370727539,
3589
+ "loss": 0.3432,
3590
+ "rewards/accuracies": 0.65625,
3591
+ "rewards/chosen": 0.7829986214637756,
3592
+ "rewards/margins": 2.0920679569244385,
3593
+ "rewards/rejected": -1.3090696334838867,
3594
+ "step": 478
3595
+ },
3596
+ {
3597
+ "epoch": 0.5972470643129326,
3598
+ "grad_norm": 23.452205657958984,
3599
+ "learning_rate": 1.701925267618779e-07,
3600
+ "logits/chosen": -10.306497573852539,
3601
+ "logits/rejected": -10.301539421081543,
3602
+ "logps/chosen": -12.428094863891602,
3603
+ "logps/rejected": -42.2680778503418,
3604
+ "loss": 0.3443,
3605
+ "rewards/accuracies": 0.84375,
3606
+ "rewards/chosen": 0.7420932054519653,
3607
+ "rewards/margins": 2.870239734649658,
3608
+ "rewards/rejected": -2.1281466484069824,
3609
+ "step": 480
3610
+ },
3611
+ {
3612
+ "epoch": 0.5997355937475698,
3613
+ "grad_norm": 17.584455490112305,
3614
+ "learning_rate": 1.6989475852277648e-07,
3615
+ "logits/chosen": -10.365903854370117,
3616
+ "logits/rejected": -10.36944580078125,
3617
+ "logps/chosen": -12.696708679199219,
3618
+ "logps/rejected": -32.62185287475586,
3619
+ "loss": 0.4025,
3620
+ "rewards/accuracies": 0.65625,
3621
+ "rewards/chosen": 0.6551677584648132,
3622
+ "rewards/margins": 1.8221435546875,
3623
+ "rewards/rejected": -1.1669756174087524,
3624
+ "step": 482
3625
+ },
3626
+ {
3627
+ "epoch": 0.602224123182207,
3628
+ "grad_norm": 10.708189964294434,
3629
+ "learning_rate": 1.6959577366697987e-07,
3630
+ "logits/chosen": -10.35180950164795,
3631
+ "logits/rejected": -10.352283477783203,
3632
+ "logps/chosen": -12.789751052856445,
3633
+ "logps/rejected": -31.80221176147461,
3634
+ "loss": 0.4418,
3635
+ "rewards/accuracies": 0.59375,
3636
+ "rewards/chosen": 0.6862015128135681,
3637
+ "rewards/margins": 1.8107177019119263,
3638
+ "rewards/rejected": -1.1245163679122925,
3639
+ "step": 484
3640
+ },
3641
+ {
3642
+ "epoch": 0.6047126526168443,
3643
+ "grad_norm": 82.61125946044922,
3644
+ "learning_rate": 1.692955773987406e-07,
3645
+ "logits/chosen": -10.326025009155273,
3646
+ "logits/rejected": -10.33124828338623,
3647
+ "logps/chosen": -16.70273780822754,
3648
+ "logps/rejected": -36.63812255859375,
3649
+ "loss": 0.3382,
3650
+ "rewards/accuracies": 0.625,
3651
+ "rewards/chosen": 0.31259140372276306,
3652
+ "rewards/margins": 1.9043779373168945,
3653
+ "rewards/rejected": -1.5917866230010986,
3654
+ "step": 486
3655
+ },
3656
+ {
3657
+ "epoch": 0.6072011820514814,
3658
+ "grad_norm": 34.93560791015625,
3659
+ "learning_rate": 1.6899417494339735e-07,
3660
+ "logits/chosen": -10.35297679901123,
3661
+ "logits/rejected": -10.35146713256836,
3662
+ "logps/chosen": -16.83978843688965,
3663
+ "logps/rejected": -41.219154357910156,
3664
+ "loss": 0.4275,
3665
+ "rewards/accuracies": 0.78125,
3666
+ "rewards/chosen": 0.27934500575065613,
3667
+ "rewards/margins": 2.2268595695495605,
3668
+ "rewards/rejected": -1.947514533996582,
3669
+ "step": 488
3670
+ },
3671
+ {
3672
+ "epoch": 0.6096897114861187,
3673
+ "grad_norm": 26.953933715820312,
3674
+ "learning_rate": 1.6869157154728434e-07,
3675
+ "logits/chosen": -10.358343124389648,
3676
+ "logits/rejected": -10.357391357421875,
3677
+ "logps/chosen": -16.978187561035156,
3678
+ "logps/rejected": -39.387386322021484,
3679
+ "loss": 0.3887,
3680
+ "rewards/accuracies": 0.78125,
3681
+ "rewards/chosen": 0.2855275273323059,
3682
+ "rewards/margins": 2.0412650108337402,
3683
+ "rewards/rejected": -1.7557373046875,
3684
+ "step": 490
3685
+ },
3686
+ {
3687
+ "epoch": 0.6121782409207559,
3688
+ "grad_norm": 24.232486724853516,
3689
+ "learning_rate": 1.6838777247763977e-07,
3690
+ "logits/chosen": -10.354990005493164,
3691
+ "logits/rejected": -10.35480785369873,
3692
+ "logps/chosen": -18.894657135009766,
3693
+ "logps/rejected": -42.638187408447266,
3694
+ "loss": 0.4061,
3695
+ "rewards/accuracies": 0.78125,
3696
+ "rewards/chosen": 0.10340137779712677,
3697
+ "rewards/margins": 2.2326226234436035,
3698
+ "rewards/rejected": -2.1292214393615723,
3699
+ "step": 492
3700
+ },
3701
+ {
3702
+ "epoch": 0.6146667703553931,
3703
+ "grad_norm": 11.788183212280273,
3704
+ "learning_rate": 1.680827830225142e-07,
3705
+ "logits/chosen": -10.291640281677246,
3706
+ "logits/rejected": -10.290667533874512,
3707
+ "logps/chosen": -16.877883911132812,
3708
+ "logps/rejected": -37.31455993652344,
3709
+ "loss": 0.3598,
3710
+ "rewards/accuracies": 0.5625,
3711
+ "rewards/chosen": 0.24069812893867493,
3712
+ "rewards/margins": 1.9603172540664673,
3713
+ "rewards/rejected": -1.7196190357208252,
3714
+ "step": 494
3715
+ },
3716
+ {
3717
+ "epoch": 0.6171552997900304,
3718
+ "grad_norm": 62.49164581298828,
3719
+ "learning_rate": 1.6777660849067868e-07,
3720
+ "logits/chosen": -10.331537246704102,
3721
+ "logits/rejected": -10.322046279907227,
3722
+ "logps/chosen": -17.917110443115234,
3723
+ "logps/rejected": -47.49253845214844,
3724
+ "loss": 0.3699,
3725
+ "rewards/accuracies": 0.78125,
3726
+ "rewards/chosen": 0.2798649072647095,
3727
+ "rewards/margins": 2.886807918548584,
3728
+ "rewards/rejected": -2.606943130493164,
3729
+ "step": 496
3730
+ },
3731
+ {
3732
+ "epoch": 0.6196438292246675,
3733
+ "grad_norm": 28.459163665771484,
3734
+ "learning_rate": 1.6746925421153194e-07,
3735
+ "logits/chosen": -10.349085807800293,
3736
+ "logits/rejected": -10.348950386047363,
3737
+ "logps/chosen": -13.380606651306152,
3738
+ "logps/rejected": -47.913330078125,
3739
+ "loss": 0.3582,
3740
+ "rewards/accuracies": 0.65625,
3741
+ "rewards/chosen": 0.6122135519981384,
3742
+ "rewards/margins": 3.276073694229126,
3743
+ "rewards/rejected": -2.6638600826263428,
3744
+ "step": 498
3745
+ },
3746
+ {
3747
+ "epoch": 0.6221323586593047,
3748
+ "grad_norm": 24.016551971435547,
3749
+ "learning_rate": 1.6716072553500812e-07,
3750
+ "logits/chosen": -10.327668190002441,
3751
+ "logits/rejected": -10.319803237915039,
3752
+ "logps/chosen": -15.610799789428711,
3753
+ "logps/rejected": -47.097747802734375,
3754
+ "loss": 0.354,
3755
+ "rewards/accuracies": 0.6875,
3756
+ "rewards/chosen": 0.42508575320243835,
3757
+ "rewards/margins": 2.950751781463623,
3758
+ "rewards/rejected": -2.5256659984588623,
3759
+ "step": 500
3760
+ },
3761
+ {
3762
+ "epoch": 0.624620888093942,
3763
+ "grad_norm": 20.961679458618164,
3764
+ "learning_rate": 1.6685102783148328e-07,
3765
+ "logits/chosen": -10.36239242553711,
3766
+ "logits/rejected": -10.371294975280762,
3767
+ "logps/chosen": -15.701215744018555,
3768
+ "logps/rejected": -39.55763244628906,
3769
+ "loss": 0.3573,
3770
+ "rewards/accuracies": 0.59375,
3771
+ "rewards/chosen": 0.39642098546028137,
3772
+ "rewards/margins": 2.26676082611084,
3773
+ "rewards/rejected": -1.8703395128250122,
3774
+ "step": 502
3775
+ },
3776
+ {
3777
+ "epoch": 0.6271094175285792,
3778
+ "grad_norm": 17.57459259033203,
3779
+ "learning_rate": 1.66540166491682e-07,
3780
+ "logits/chosen": -10.354241371154785,
3781
+ "logits/rejected": -10.34440803527832,
3782
+ "logps/chosen": -14.872220993041992,
3783
+ "logps/rejected": -38.538856506347656,
3784
+ "loss": 0.3463,
3785
+ "rewards/accuracies": 0.6875,
3786
+ "rewards/chosen": 0.47063276171684265,
3787
+ "rewards/margins": 2.1610333919525146,
3788
+ "rewards/rejected": -1.6904008388519287,
3789
+ "step": 504
3790
+ },
3791
+ {
3792
+ "epoch": 0.6295979469632165,
3793
+ "grad_norm": 60.06468200683594,
3794
+ "learning_rate": 1.6622814692658369e-07,
3795
+ "logits/chosen": -10.337247848510742,
3796
+ "logits/rejected": -10.33508586883545,
3797
+ "logps/chosen": -17.578996658325195,
3798
+ "logps/rejected": -37.28183364868164,
3799
+ "loss": 0.3645,
3800
+ "rewards/accuracies": 0.65625,
3801
+ "rewards/chosen": 0.21325191855430603,
3802
+ "rewards/margins": 1.7064669132232666,
3803
+ "rewards/rejected": -1.4932150840759277,
3804
+ "step": 506
3805
+ },
3806
+ {
3807
+ "epoch": 0.6320864763978536,
3808
+ "grad_norm": 135.35751342773438,
3809
+ "learning_rate": 1.6591497456732824e-07,
3810
+ "logits/chosen": -10.389504432678223,
3811
+ "logits/rejected": -10.391674041748047,
3812
+ "logps/chosen": -18.306015014648438,
3813
+ "logps/rejected": -48.028743743896484,
3814
+ "loss": 0.355,
3815
+ "rewards/accuracies": 0.75,
3816
+ "rewards/chosen": 0.18720369040966034,
3817
+ "rewards/margins": 2.9284324645996094,
3818
+ "rewards/rejected": -2.7412288188934326,
3819
+ "step": 508
3820
+ },
3821
+ {
3822
+ "epoch": 0.6345750058324908,
3823
+ "grad_norm": 87.70523071289062,
3824
+ "learning_rate": 1.6560065486512156e-07,
3825
+ "logits/chosen": -10.294554710388184,
3826
+ "logits/rejected": -10.290327072143555,
3827
+ "logps/chosen": -18.5239315032959,
3828
+ "logps/rejected": -45.670509338378906,
3829
+ "loss": 0.3753,
3830
+ "rewards/accuracies": 0.625,
3831
+ "rewards/chosen": 0.16150438785552979,
3832
+ "rewards/margins": 2.575958490371704,
3833
+ "rewards/rejected": -2.4144539833068848,
3834
+ "step": 510
3835
+ },
3836
+ {
3837
+ "epoch": 0.6370635352671281,
3838
+ "grad_norm": 46.76156234741211,
3839
+ "learning_rate": 1.6528519329114068e-07,
3840
+ "logits/chosen": -10.382038116455078,
3841
+ "logits/rejected": -10.384214401245117,
3842
+ "logps/chosen": -21.97479248046875,
3843
+ "logps/rejected": -51.65577697753906,
3844
+ "loss": 0.3262,
3845
+ "rewards/accuracies": 0.65625,
3846
+ "rewards/chosen": -0.26933473348617554,
3847
+ "rewards/margins": 2.7787156105041504,
3848
+ "rewards/rejected": -3.0480501651763916,
3849
+ "step": 512
3850
+ },
3851
+ {
3852
+ "epoch": 0.6395520647017653,
3853
+ "grad_norm": 31.331327438354492,
3854
+ "learning_rate": 1.649685953364385e-07,
3855
+ "logits/chosen": -10.314241409301758,
3856
+ "logits/rejected": -10.318497657775879,
3857
+ "logps/chosen": -20.867794036865234,
3858
+ "logps/rejected": -42.47340393066406,
3859
+ "loss": 0.3523,
3860
+ "rewards/accuracies": 0.59375,
3861
+ "rewards/chosen": -0.10111423581838608,
3862
+ "rewards/margins": 2.1302971839904785,
3863
+ "rewards/rejected": -2.2314112186431885,
3864
+ "step": 514
3865
+ },
3866
+ {
3867
+ "epoch": 0.6420405941364025,
3868
+ "grad_norm": 14.808184623718262,
3869
+ "learning_rate": 1.6465086651184825e-07,
3870
+ "logits/chosen": -10.327038764953613,
3871
+ "logits/rejected": -10.305490493774414,
3872
+ "logps/chosen": -20.10313606262207,
3873
+ "logps/rejected": -42.85016632080078,
3874
+ "loss": 0.3234,
3875
+ "rewards/accuracies": 0.625,
3876
+ "rewards/chosen": -0.052409958094358444,
3877
+ "rewards/margins": 2.088395833969116,
3878
+ "rewards/rejected": -2.140805721282959,
3879
+ "step": 516
3880
+ },
3881
+ {
3882
+ "epoch": 0.6445291235710398,
3883
+ "grad_norm": 93.76280975341797,
3884
+ "learning_rate": 1.6433201234788756e-07,
3885
+ "logits/chosen": -10.36878490447998,
3886
+ "logits/rejected": -10.36519718170166,
3887
+ "logps/chosen": -19.352113723754883,
3888
+ "logps/rejected": -40.40311050415039,
3889
+ "loss": 0.3616,
3890
+ "rewards/accuracies": 0.71875,
3891
+ "rewards/chosen": 0.03417102247476578,
3892
+ "rewards/margins": 1.9921482801437378,
3893
+ "rewards/rejected": -1.957977056503296,
3894
+ "step": 518
3895
+ },
3896
+ {
3897
+ "epoch": 0.6470176530056769,
3898
+ "grad_norm": 56.11380386352539,
3899
+ "learning_rate": 1.6401203839466212e-07,
3900
+ "logits/chosen": -10.365287780761719,
3901
+ "logits/rejected": -10.368802070617676,
3902
+ "logps/chosen": -25.988582611083984,
3903
+ "logps/rejected": -59.596805572509766,
3904
+ "loss": 0.3922,
3905
+ "rewards/accuracies": 0.8125,
3906
+ "rewards/chosen": -0.6514653563499451,
3907
+ "rewards/margins": 3.0986266136169434,
3908
+ "rewards/rejected": -3.7500922679901123,
3909
+ "step": 520
3910
+ },
3911
+ {
3912
+ "epoch": 0.6495061824403142,
3913
+ "grad_norm": 9.446833610534668,
3914
+ "learning_rate": 1.6369095022176917e-07,
3915
+ "logits/chosen": -10.37309741973877,
3916
+ "logits/rejected": -10.370424270629883,
3917
+ "logps/chosen": -20.687456130981445,
3918
+ "logps/rejected": -54.72138595581055,
3919
+ "loss": 0.3037,
3920
+ "rewards/accuracies": 0.75,
3921
+ "rewards/chosen": -0.1104515939950943,
3922
+ "rewards/margins": 3.277845859527588,
3923
+ "rewards/rejected": -3.3882977962493896,
3924
+ "step": 522
3925
+ },
3926
+ {
3927
+ "epoch": 0.6519947118749514,
3928
+ "grad_norm": 35.417701721191406,
3929
+ "learning_rate": 1.6336875341820048e-07,
3930
+ "logits/chosen": -10.3366060256958,
3931
+ "logits/rejected": -10.33828067779541,
3932
+ "logps/chosen": -16.212318420410156,
3933
+ "logps/rejected": -43.318729400634766,
3934
+ "loss": 0.2952,
3935
+ "rewards/accuracies": 0.6875,
3936
+ "rewards/chosen": 0.3492877185344696,
3937
+ "rewards/margins": 2.614508867263794,
3938
+ "rewards/rejected": -2.265221118927002,
3939
+ "step": 524
3940
+ },
3941
+ {
3942
+ "epoch": 0.6544832413095886,
3943
+ "grad_norm": 12.384411811828613,
3944
+ "learning_rate": 1.630454535922452e-07,
3945
+ "logits/chosen": -10.360422134399414,
3946
+ "logits/rejected": -10.380635261535645,
3947
+ "logps/chosen": -14.600409507751465,
3948
+ "logps/rejected": -46.50564956665039,
3949
+ "loss": 0.3465,
3950
+ "rewards/accuracies": 0.65625,
3951
+ "rewards/chosen": 0.5301249027252197,
3952
+ "rewards/margins": 3.087332010269165,
3953
+ "rewards/rejected": -2.5572071075439453,
3954
+ "step": 526
3955
+ },
3956
+ {
3957
+ "epoch": 0.6569717707442259,
3958
+ "grad_norm": 18.790855407714844,
3959
+ "learning_rate": 1.62721056371392e-07,
3960
+ "logits/chosen": -10.37255859375,
3961
+ "logits/rejected": -10.366708755493164,
3962
+ "logps/chosen": -16.081214904785156,
3963
+ "logps/rejected": -53.83919906616211,
3964
+ "loss": 0.2977,
3965
+ "rewards/accuracies": 0.75,
3966
+ "rewards/chosen": 0.3910747468471527,
3967
+ "rewards/margins": 3.5673904418945312,
3968
+ "rewards/rejected": -3.1763155460357666,
3969
+ "step": 528
3970
+ },
3971
+ {
3972
+ "epoch": 0.659460300178863,
3973
+ "grad_norm": 22.111038208007812,
3974
+ "learning_rate": 1.6239556740223131e-07,
3975
+ "logits/chosen": -10.393640518188477,
3976
+ "logits/rejected": -10.395002365112305,
3977
+ "logps/chosen": -13.953643798828125,
3978
+ "logps/rejected": -36.29132080078125,
3979
+ "loss": 0.3785,
3980
+ "rewards/accuracies": 0.6875,
3981
+ "rewards/chosen": 0.5791424512863159,
3982
+ "rewards/margins": 2.1625773906707764,
3983
+ "rewards/rejected": -1.58343505859375,
3984
+ "step": 530
3985
+ },
3986
+ {
3987
+ "epoch": 0.6619488296135003,
3988
+ "grad_norm": 6.779727458953857,
3989
+ "learning_rate": 1.62068992350357e-07,
3990
+ "logits/chosen": -10.42313003540039,
3991
+ "logits/rejected": -10.428924560546875,
3992
+ "logps/chosen": -13.445175170898438,
3993
+ "logps/rejected": -53.513519287109375,
3994
+ "loss": 0.2684,
3995
+ "rewards/accuracies": 0.6875,
3996
+ "rewards/chosen": 0.643733024597168,
3997
+ "rewards/margins": 3.7983388900756836,
3998
+ "rewards/rejected": -3.1546058654785156,
3999
+ "step": 532
4000
+ },
4001
+ {
4002
+ "epoch": 0.6644373590481375,
4003
+ "grad_norm": 24.651657104492188,
4004
+ "learning_rate": 1.6174133690026766e-07,
4005
+ "logits/chosen": -10.423484802246094,
4006
+ "logits/rejected": -10.423924446105957,
4007
+ "logps/chosen": -17.091449737548828,
4008
+ "logps/rejected": -46.30265808105469,
4009
+ "loss": 0.3557,
4010
+ "rewards/accuracies": 0.75,
4011
+ "rewards/chosen": 0.28270694613456726,
4012
+ "rewards/margins": 2.6901934146881104,
4013
+ "rewards/rejected": -2.4074862003326416,
4014
+ "step": 534
4015
+ },
4016
+ {
4017
+ "epoch": 0.6669258884827747,
4018
+ "grad_norm": 12.76252555847168,
4019
+ "learning_rate": 1.614126067552679e-07,
4020
+ "logits/chosen": -10.382489204406738,
4021
+ "logits/rejected": -10.38238525390625,
4022
+ "logps/chosen": -17.7365779876709,
4023
+ "logps/rejected": -43.99931716918945,
4024
+ "loss": 0.3144,
4025
+ "rewards/accuracies": 0.65625,
4026
+ "rewards/chosen": 0.1616898477077484,
4027
+ "rewards/margins": 2.424696922302246,
4028
+ "rewards/rejected": -2.2630069255828857,
4029
+ "step": 536
4030
+ },
4031
+ {
4032
+ "epoch": 0.669414417917412,
4033
+ "grad_norm": 14.746679306030273,
4034
+ "learning_rate": 1.610828076373687e-07,
4035
+ "logits/chosen": -10.409899711608887,
4036
+ "logits/rejected": -10.403095245361328,
4037
+ "logps/chosen": -19.982406616210938,
4038
+ "logps/rejected": -53.89605712890625,
4039
+ "loss": 0.381,
4040
+ "rewards/accuracies": 0.65625,
4041
+ "rewards/chosen": -0.03805060684680939,
4042
+ "rewards/margins": 3.2466320991516113,
4043
+ "rewards/rejected": -3.2846827507019043,
4044
+ "step": 538
4045
+ },
4046
+ {
4047
+ "epoch": 0.6719029473520491,
4048
+ "grad_norm": 28.71502685546875,
4049
+ "learning_rate": 1.6075194528718816e-07,
4050
+ "logits/chosen": -10.433353424072266,
4051
+ "logits/rejected": -10.431252479553223,
4052
+ "logps/chosen": -20.391389846801758,
4053
+ "logps/rejected": -46.728179931640625,
4054
+ "loss": 0.4031,
4055
+ "rewards/accuracies": 0.75,
4056
+ "rewards/chosen": -0.08024211972951889,
4057
+ "rewards/margins": 2.4565019607543945,
4058
+ "rewards/rejected": -2.5367441177368164,
4059
+ "step": 540
4060
+ },
4061
+ {
4062
+ "epoch": 0.6743914767866863,
4063
+ "grad_norm": 75.32211303710938,
4064
+ "learning_rate": 1.6042002546385138e-07,
4065
+ "logits/chosen": -10.414860725402832,
4066
+ "logits/rejected": -10.422852516174316,
4067
+ "logps/chosen": -22.48684310913086,
4068
+ "logps/rejected": -55.422760009765625,
4069
+ "loss": 0.3963,
4070
+ "rewards/accuracies": 0.6875,
4071
+ "rewards/chosen": -0.26822561025619507,
4072
+ "rewards/margins": 3.1868667602539062,
4073
+ "rewards/rejected": -3.455092430114746,
4074
+ "step": 542
4075
+ },
4076
+ {
4077
+ "epoch": 0.6768800062213236,
4078
+ "grad_norm": 22.97120475769043,
4079
+ "learning_rate": 1.600870539448903e-07,
4080
+ "logits/chosen": -10.432124137878418,
4081
+ "logits/rejected": -10.43125057220459,
4082
+ "logps/chosen": -21.589643478393555,
4083
+ "logps/rejected": -52.47227096557617,
4084
+ "loss": 0.4153,
4085
+ "rewards/accuracies": 0.625,
4086
+ "rewards/chosen": -0.16771256923675537,
4087
+ "rewards/margins": 2.870025157928467,
4088
+ "rewards/rejected": -3.0377376079559326,
4089
+ "step": 544
4090
+ },
4091
+ {
4092
+ "epoch": 0.6793685356559608,
4093
+ "grad_norm": 11.576513290405273,
4094
+ "learning_rate": 1.5975303652614308e-07,
4095
+ "logits/chosen": -10.42074203491211,
4096
+ "logits/rejected": -10.41093921661377,
4097
+ "logps/chosen": -20.702247619628906,
4098
+ "logps/rejected": -45.585052490234375,
4099
+ "loss": 0.3736,
4100
+ "rewards/accuracies": 0.75,
4101
+ "rewards/chosen": -0.11785610765218735,
4102
+ "rewards/margins": 2.25642991065979,
4103
+ "rewards/rejected": -2.37428617477417,
4104
+ "step": 546
4105
+ },
4106
+ {
4107
+ "epoch": 0.6818570650905981,
4108
+ "grad_norm": 38.1824836730957,
4109
+ "learning_rate": 1.5941797902165321e-07,
4110
+ "logits/chosen": -10.406185150146484,
4111
+ "logits/rejected": -10.406208038330078,
4112
+ "logps/chosen": -17.893470764160156,
4113
+ "logps/rejected": -47.94472885131836,
4114
+ "loss": 0.3662,
4115
+ "rewards/accuracies": 0.6875,
4116
+ "rewards/chosen": 0.22653920948505402,
4117
+ "rewards/margins": 2.9794092178344727,
4118
+ "rewards/rejected": -2.7528700828552246,
4119
+ "step": 548
4120
+ },
4121
+ {
4122
+ "epoch": 0.6843455945252352,
4123
+ "grad_norm": 8.212292671203613,
4124
+ "learning_rate": 1.5908188726356842e-07,
4125
+ "logits/chosen": -10.435319900512695,
4126
+ "logits/rejected": -10.433416366577148,
4127
+ "logps/chosen": -15.427078247070312,
4128
+ "logps/rejected": -37.61276626586914,
4129
+ "loss": 0.3822,
4130
+ "rewards/accuracies": 0.625,
4131
+ "rewards/chosen": 0.4034353792667389,
4132
+ "rewards/margins": 2.0808515548706055,
4133
+ "rewards/rejected": -1.6774165630340576,
4134
+ "step": 550
4135
+ },
4136
+ {
4137
+ "epoch": 0.6868341239598724,
4138
+ "grad_norm": 7.298886299133301,
4139
+ "learning_rate": 1.58744767102039e-07,
4140
+ "logits/chosen": -10.464427947998047,
4141
+ "logits/rejected": -10.468774795532227,
4142
+ "logps/chosen": -14.835740089416504,
4143
+ "logps/rejected": -43.39911651611328,
4144
+ "loss": 0.3506,
4145
+ "rewards/accuracies": 0.6875,
4146
+ "rewards/chosen": 0.4960540235042572,
4147
+ "rewards/margins": 2.7094733715057373,
4148
+ "rewards/rejected": -2.213419198989868,
4149
+ "step": 552
4150
+ },
4151
+ {
4152
+ "epoch": 0.6893226533945097,
4153
+ "grad_norm": 11.580885887145996,
4154
+ "learning_rate": 1.5840662440511604e-07,
4155
+ "logits/chosen": -10.460016250610352,
4156
+ "logits/rejected": -10.45378303527832,
4157
+ "logps/chosen": -7.484930992126465,
4158
+ "logps/rejected": -45.44110107421875,
4159
+ "loss": 0.2906,
4160
+ "rewards/accuracies": 0.71875,
4161
+ "rewards/chosen": 1.189929723739624,
4162
+ "rewards/margins": 3.5907058715820312,
4163
+ "rewards/rejected": -2.400775909423828,
4164
+ "step": 554
4165
+ },
4166
+ {
4167
+ "epoch": 0.6918111828291469,
4168
+ "grad_norm": 10.710088729858398,
4169
+ "learning_rate": 1.5806746505864946e-07,
4170
+ "logits/chosen": -10.52313232421875,
4171
+ "logits/rejected": -10.540308952331543,
4172
+ "logps/chosen": -8.493289947509766,
4173
+ "logps/rejected": -57.8527946472168,
4174
+ "loss": 0.2883,
4175
+ "rewards/accuracies": 0.78125,
4176
+ "rewards/chosen": 1.0914571285247803,
4177
+ "rewards/margins": 4.764283180236816,
4178
+ "rewards/rejected": -3.672825813293457,
4179
+ "step": 556
4180
+ },
4181
+ {
4182
+ "epoch": 0.6942997122637842,
4183
+ "grad_norm": 11.217126846313477,
4184
+ "learning_rate": 1.5772729496618515e-07,
4185
+ "logits/chosen": -10.509547233581543,
4186
+ "logits/rejected": -10.509346008300781,
4187
+ "logps/chosen": -13.166345596313477,
4188
+ "logps/rejected": -25.258852005004883,
4189
+ "loss": 0.3988,
4190
+ "rewards/accuracies": 0.4375,
4191
+ "rewards/chosen": 0.6617048382759094,
4192
+ "rewards/margins": 1.1463922262191772,
4193
+ "rewards/rejected": -0.4846874475479126,
4194
+ "step": 558
4195
+ },
4196
+ {
4197
+ "epoch": 0.6967882416984214,
4198
+ "grad_norm": 81.53024291992188,
4199
+ "learning_rate": 1.5738612004886265e-07,
4200
+ "logits/chosen": -10.518564224243164,
4201
+ "logits/rejected": -10.512411117553711,
4202
+ "logps/chosen": -6.1867218017578125,
4203
+ "logps/rejected": -28.618839263916016,
4204
+ "loss": 0.3907,
4205
+ "rewards/accuracies": 0.71875,
4206
+ "rewards/chosen": 1.327723503112793,
4207
+ "rewards/margins": 1.9867522716522217,
4208
+ "rewards/rejected": -0.6590288877487183,
4209
+ "step": 560
4210
+ },
4211
+ {
4212
+ "epoch": 0.6992767711330585,
4213
+ "grad_norm": 15.888360023498535,
4214
+ "learning_rate": 1.5704394624531183e-07,
4215
+ "logits/chosen": -10.48656177520752,
4216
+ "logits/rejected": -10.485774993896484,
4217
+ "logps/chosen": -5.765016078948975,
4218
+ "logps/rejected": -26.678476333618164,
4219
+ "loss": 0.3037,
4220
+ "rewards/accuracies": 0.5625,
4221
+ "rewards/chosen": 1.4207212924957275,
4222
+ "rewards/margins": 1.9595447778701782,
4223
+ "rewards/rejected": -0.5388234853744507,
4224
+ "step": 562
4225
+ },
4226
+ {
4227
+ "epoch": 0.7017653005676958,
4228
+ "grad_norm": 97.72893524169922,
4229
+ "learning_rate": 1.5670077951154955e-07,
4230
+ "logits/chosen": -10.544830322265625,
4231
+ "logits/rejected": -10.549147605895996,
4232
+ "logps/chosen": -10.575237274169922,
4233
+ "logps/rejected": -34.218902587890625,
4234
+ "loss": 0.3856,
4235
+ "rewards/accuracies": 0.65625,
4236
+ "rewards/chosen": 0.9205721020698547,
4237
+ "rewards/margins": 2.3256759643554688,
4238
+ "rewards/rejected": -1.4051040410995483,
4239
+ "step": 564
4240
+ },
4241
+ {
4242
+ "epoch": 0.704253830002333,
4243
+ "grad_norm": 26.422163009643555,
4244
+ "learning_rate": 1.5635662582087603e-07,
4245
+ "logits/chosen": -10.535964965820312,
4246
+ "logits/rejected": -10.541629791259766,
4247
+ "logps/chosen": -11.3693265914917,
4248
+ "logps/rejected": -43.26926040649414,
4249
+ "loss": 0.374,
4250
+ "rewards/accuracies": 0.625,
4251
+ "rewards/chosen": 0.8099326491355896,
4252
+ "rewards/margins": 3.01070237159729,
4253
+ "rewards/rejected": -2.2007696628570557,
4254
+ "step": 566
4255
+ },
4256
+ {
4257
+ "epoch": 0.7067423594369702,
4258
+ "grad_norm": 33.08991241455078,
4259
+ "learning_rate": 1.5601149116377092e-07,
4260
+ "logits/chosen": -10.547080039978027,
4261
+ "logits/rejected": -10.545144081115723,
4262
+ "logps/chosen": -11.223136901855469,
4263
+ "logps/rejected": -32.360145568847656,
4264
+ "loss": 0.3539,
4265
+ "rewards/accuracies": 0.71875,
4266
+ "rewards/chosen": 0.824683427810669,
4267
+ "rewards/margins": 1.9326567649841309,
4268
+ "rewards/rejected": -1.107973337173462,
4269
+ "step": 568
4270
+ },
4271
+ {
4272
+ "epoch": 0.7092308888716075,
4273
+ "grad_norm": 4.160190105438232,
4274
+ "learning_rate": 1.5566538154778893e-07,
4275
+ "logits/chosen": -10.599503517150879,
4276
+ "logits/rejected": -10.586050033569336,
4277
+ "logps/chosen": -12.865560531616211,
4278
+ "logps/rejected": -47.178524017333984,
4279
+ "loss": 0.2787,
4280
+ "rewards/accuracies": 0.71875,
4281
+ "rewards/chosen": 0.6689233779907227,
4282
+ "rewards/margins": 3.2086825370788574,
4283
+ "rewards/rejected": -2.539759397506714,
4284
+ "step": 570
4285
+ },
4286
+ {
4287
+ "epoch": 0.7117194183062446,
4288
+ "grad_norm": 10.538558006286621,
4289
+ "learning_rate": 1.5531830299745528e-07,
4290
+ "logits/chosen": -10.571596145629883,
4291
+ "logits/rejected": -10.56047534942627,
4292
+ "logps/chosen": -10.884844779968262,
4293
+ "logps/rejected": -39.770355224609375,
4294
+ "loss": 0.3544,
4295
+ "rewards/accuracies": 0.59375,
4296
+ "rewards/chosen": 0.8782658576965332,
4297
+ "rewards/margins": 2.7163915634155273,
4298
+ "rewards/rejected": -1.8381257057189941,
4299
+ "step": 572
4300
+ },
4301
+ {
4302
+ "epoch": 0.7142079477408819,
4303
+ "grad_norm": 103.00370788574219,
4304
+ "learning_rate": 1.5497026155416086e-07,
4305
+ "logits/chosen": -10.592615127563477,
4306
+ "logits/rejected": -10.590554237365723,
4307
+ "logps/chosen": -13.306787490844727,
4308
+ "logps/rejected": -47.153350830078125,
4309
+ "loss": 0.3521,
4310
+ "rewards/accuracies": 0.65625,
4311
+ "rewards/chosen": 0.6575937867164612,
4312
+ "rewards/margins": 3.320206642150879,
4313
+ "rewards/rejected": -2.6626124382019043,
4314
+ "step": 574
4315
+ },
4316
+ {
4317
+ "epoch": 0.7166964771755191,
4318
+ "grad_norm": 7.326557636260986,
4319
+ "learning_rate": 1.5462126327605716e-07,
4320
+ "logits/chosen": -10.630630493164062,
4321
+ "logits/rejected": -10.631823539733887,
4322
+ "logps/chosen": -15.150641441345215,
4323
+ "logps/rejected": -41.585872650146484,
4324
+ "loss": 0.3273,
4325
+ "rewards/accuracies": 0.75,
4326
+ "rewards/chosen": 0.4338640868663788,
4327
+ "rewards/margins": 2.4150288105010986,
4328
+ "rewards/rejected": -1.9811646938323975,
4329
+ "step": 576
4330
+ },
4331
+ {
4332
+ "epoch": 0.7191850066101563,
4333
+ "grad_norm": 6.447935104370117,
4334
+ "learning_rate": 1.5427131423795058e-07,
4335
+ "logits/chosen": -10.528682708740234,
4336
+ "logits/rejected": -10.520992279052734,
4337
+ "logps/chosen": -13.810688972473145,
4338
+ "logps/rejected": -40.82366943359375,
4339
+ "loss": 0.342,
4340
+ "rewards/accuracies": 0.59375,
4341
+ "rewards/chosen": 0.5959270596504211,
4342
+ "rewards/margins": 2.6108908653259277,
4343
+ "rewards/rejected": -2.0149636268615723,
4344
+ "step": 578
4345
+ },
4346
+ {
4347
+ "epoch": 0.7216735360447936,
4348
+ "grad_norm": 47.59484100341797,
4349
+ "learning_rate": 1.5392042053119697e-07,
4350
+ "logits/chosen": -10.53530502319336,
4351
+ "logits/rejected": -10.529587745666504,
4352
+ "logps/chosen": -14.466578483581543,
4353
+ "logps/rejected": -35.32248306274414,
4354
+ "loss": 0.4071,
4355
+ "rewards/accuracies": 0.625,
4356
+ "rewards/chosen": 0.52098149061203,
4357
+ "rewards/margins": 1.9215223789215088,
4358
+ "rewards/rejected": -1.4005409479141235,
4359
+ "step": 580
4360
+ },
4361
+ {
4362
+ "epoch": 0.7241620654794307,
4363
+ "grad_norm": 47.81111526489258,
4364
+ "learning_rate": 1.535685882635954e-07,
4365
+ "logits/chosen": -10.54077434539795,
4366
+ "logits/rejected": -10.534811973571777,
4367
+ "logps/chosen": -14.007987976074219,
4368
+ "logps/rejected": -31.1512451171875,
4369
+ "loss": 0.4098,
4370
+ "rewards/accuracies": 0.53125,
4371
+ "rewards/chosen": 0.5760743021965027,
4372
+ "rewards/margins": 1.6659516096115112,
4373
+ "rewards/rejected": -1.0898771286010742,
4374
+ "step": 582
4375
+ },
4376
+ {
4377
+ "epoch": 0.7266505949140679,
4378
+ "grad_norm": 12.10058307647705,
4379
+ "learning_rate": 1.5321582355928189e-07,
4380
+ "logits/chosen": -10.575304985046387,
4381
+ "logits/rejected": -10.571596145629883,
4382
+ "logps/chosen": -13.538256645202637,
4383
+ "logps/rejected": -49.876441955566406,
4384
+ "loss": 0.3588,
4385
+ "rewards/accuracies": 0.8125,
4386
+ "rewards/chosen": 0.5914421677589417,
4387
+ "rewards/margins": 3.4231691360473633,
4388
+ "rewards/rejected": -2.8317270278930664,
4389
+ "step": 584
4390
+ },
4391
+ {
4392
+ "epoch": 0.7291391243487052,
4393
+ "grad_norm": 14.012777328491211,
4394
+ "learning_rate": 1.5286213255862293e-07,
4395
+ "logits/chosen": -10.537784576416016,
4396
+ "logits/rejected": -10.535847663879395,
4397
+ "logps/chosen": -13.171601295471191,
4398
+ "logps/rejected": -36.27528762817383,
4399
+ "loss": 0.4037,
4400
+ "rewards/accuracies": 0.5,
4401
+ "rewards/chosen": 0.6269873976707458,
4402
+ "rewards/margins": 2.104299306869507,
4403
+ "rewards/rejected": -1.4773118495941162,
4404
+ "step": 586
4405
+ },
4406
+ {
4407
+ "epoch": 0.7316276537833424,
4408
+ "grad_norm": 11.143157958984375,
4409
+ "learning_rate": 1.5250752141810837e-07,
4410
+ "logits/chosen": -10.586292266845703,
4411
+ "logits/rejected": -10.584369659423828,
4412
+ "logps/chosen": -17.792314529418945,
4413
+ "logps/rejected": -39.53071975708008,
4414
+ "loss": 0.3498,
4415
+ "rewards/accuracies": 0.65625,
4416
+ "rewards/chosen": 0.21173079311847687,
4417
+ "rewards/margins": 2.085752010345459,
4418
+ "rewards/rejected": -1.874021291732788,
4419
+ "step": 588
4420
+ },
4421
+ {
4422
+ "epoch": 0.7341161832179797,
4423
+ "grad_norm": 3.511929988861084,
4424
+ "learning_rate": 1.521519963102445e-07,
4425
+ "logits/chosen": -10.524730682373047,
4426
+ "logits/rejected": -10.530563354492188,
4427
+ "logps/chosen": -17.697795867919922,
4428
+ "logps/rejected": -39.3559684753418,
4429
+ "loss": 0.3913,
4430
+ "rewards/accuracies": 0.53125,
4431
+ "rewards/chosen": 0.1921282708644867,
4432
+ "rewards/margins": 1.9607605934143066,
4433
+ "rewards/rejected": -1.768632173538208,
4434
+ "step": 590
4435
+ },
4436
+ {
4437
+ "epoch": 0.7366047126526168,
4438
+ "grad_norm": 15.319831848144531,
4439
+ "learning_rate": 1.517955634234464e-07,
4440
+ "logits/chosen": -10.562938690185547,
4441
+ "logits/rejected": -10.557134628295898,
4442
+ "logps/chosen": -16.555072784423828,
4443
+ "logps/rejected": -34.56721878051758,
4444
+ "loss": 0.379,
4445
+ "rewards/accuracies": 0.59375,
4446
+ "rewards/chosen": 0.3209540545940399,
4447
+ "rewards/margins": 1.7064590454101562,
4448
+ "rewards/rejected": -1.3855050802230835,
4449
+ "step": 592
4450
+ },
4451
+ {
4452
+ "epoch": 0.739093242087254,
4453
+ "grad_norm": 49.82108688354492,
4454
+ "learning_rate": 1.514382289619305e-07,
4455
+ "logits/chosen": -10.586225509643555,
4456
+ "logits/rejected": -10.584135055541992,
4457
+ "logps/chosen": -25.68656349182129,
4458
+ "logps/rejected": -51.877830505371094,
4459
+ "loss": 0.3635,
4460
+ "rewards/accuracies": 0.53125,
4461
+ "rewards/chosen": -0.6163984537124634,
4462
+ "rewards/margins": 2.479316473007202,
4463
+ "rewards/rejected": -3.095715045928955,
4464
+ "step": 594
4465
+ },
4466
+ {
4467
+ "epoch": 0.7415817715218913,
4468
+ "grad_norm": 34.32016372680664,
4469
+ "learning_rate": 1.5107999914560615e-07,
4470
+ "logits/chosen": -10.555464744567871,
4471
+ "logits/rejected": -10.558659553527832,
4472
+ "logps/chosen": -15.712035179138184,
4473
+ "logps/rejected": -45.196128845214844,
4474
+ "loss": 0.3779,
4475
+ "rewards/accuracies": 0.59375,
4476
+ "rewards/chosen": 0.39934056997299194,
4477
+ "rewards/margins": 2.773020029067993,
4478
+ "rewards/rejected": -2.3736791610717773,
4479
+ "step": 596
4480
+ },
4481
+ {
4482
+ "epoch": 0.7440703009565285,
4483
+ "grad_norm": 72.69339752197266,
4484
+ "learning_rate": 1.507208802099679e-07,
4485
+ "logits/chosen": -10.546841621398926,
4486
+ "logits/rejected": -10.54928970336914,
4487
+ "logps/chosen": -16.846134185791016,
4488
+ "logps/rejected": -51.7169303894043,
4489
+ "loss": 0.3758,
4490
+ "rewards/accuracies": 0.75,
4491
+ "rewards/chosen": 0.2728378176689148,
4492
+ "rewards/margins": 3.2069714069366455,
4493
+ "rewards/rejected": -2.934133529663086,
4494
+ "step": 598
4495
+ },
4496
+ {
4497
+ "epoch": 0.7465588303911658,
4498
+ "grad_norm": 12.59472942352295,
4499
+ "learning_rate": 1.503608784059864e-07,
4500
+ "logits/chosen": -10.608214378356934,
4501
+ "logits/rejected": -10.605714797973633,
4502
+ "logps/chosen": -15.544697761535645,
4503
+ "logps/rejected": -43.956626892089844,
4504
+ "loss": 0.3453,
4505
+ "rewards/accuracies": 0.59375,
4506
+ "rewards/chosen": 0.37032249569892883,
4507
+ "rewards/margins": 2.7240092754364014,
4508
+ "rewards/rejected": -2.353686809539795,
4509
+ "step": 600
4510
  }
4511
  ],
4512
  "logging_steps": 2,