Training in progress, step 480, checkpoint

last-checkpoint/adapter_config.json

@@ -29,13 +29,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "up_proj",
     "gate_proj",
-    "down_proj",
-    "v_proj",
     "q_proj",
+    "down_proj",
     "k_proj",
-    "up_proj",
-    "o_proj"
+    "o_proj",
+    "v_proj"
   ],
   "task_type": "CAUSAL_LM",
   "trainable_token_indices": null,

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e97f0845aa6a67e07d4147bdbc645881cd725e485863c57bcc61520e8aa526f0
+oid sha256:493afea1d4b0a4b8fcbabd4782faf59bce83b78390d3fe22c1f7e9f9009ee34f
 size 194563400

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e076fb439be773d56148898b13cdfb393391bdee8311e86bd9314c0519333d5a
+oid sha256:8c30c5fc8db3a087f890fea74bf58a9464576de6fd24c51084c1a9e86ddbdb29
 size 100256339

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f759ddb7949658691c66ca99f40d508b0c01db148b293eb4b7d316ed5233a57e
+oid sha256:c8e1591707f95e85ddb52f431f3d9894670c8a0156ac1d112e63ad294b567127
 size 14645

last-checkpoint/scaler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8471f0babb163e30a8ef2747dbd9c4e7f32c7d3318b6dd67c171a8d396f73ab6
+oid sha256:97d84fa515775ff9ef37dff77ad8970e6a70fabe1b6fd86d0f16647a32126001
 size 1383

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b15150ddaa3faf35858f43938a74f78c04812ac2062419c41ac0afaa4a8ae794
+oid sha256:d9671054f4efb882fe928a0a76ad9e3c53090dcbe026d23891f2bbb0f73c2434
 size 1465

last-checkpoint/trainer_state.json

@@ -2,9 +2,9 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 24.736842105263158,
+  "epoch": 68.57142857142857,
   "eval_steps": 500,
-  "global_step": 470,
+  "global_step": 480,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -10348,12 +10348,232 @@
       "rewards/compute_weighted_reward/mean": 0.8570833802223206,
       "rewards/compute_weighted_reward/std": 1.384502649307251,
       "step": 470
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.16666666666666663,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1061.0,
+      "completions/mean_length": 729.3333740234375,
+      "completions/mean_terminated_length": 595.2000122070312,
+      "completions/min_length": 333.0,
+      "completions/min_terminated_length": 333.0,
+      "epoch": 67.28571428571429,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 6.319087982177734,
+      "kl": 0.008305572322569788,
+      "learning_rate": 2.642109712683971e-09,
+      "loss": 0.0083,
+      "num_tokens": 12849834.0,
+      "reward": 0.3420833945274353,
+      "reward_std": 1.2131065130233765,
+      "rewards/compute_weighted_reward/mean": 0.3420833349227905,
+      "rewards/compute_weighted_reward/std": 1.3302565813064575,
+      "step": 471
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.16666666666666663,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1239.0,
+      "completions/mean_length": 852.2916870117188,
+      "completions/mean_terminated_length": 742.75,
+      "completions/min_length": 336.0,
+      "completions/min_terminated_length": 336.0,
+      "epoch": 67.42857142857143,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 8.495345115661621,
+      "kl": 0.007552888710051775,
+      "learning_rate": 2.4850645694436735e-09,
+      "loss": 0.0076,
+      "num_tokens": 12877975.0,
+      "reward": 0.9245834350585938,
+      "reward_std": 0.9624896049499512,
+      "rewards/compute_weighted_reward/mean": 0.924583375453949,
+      "rewards/compute_weighted_reward/std": 1.2036573886871338,
+      "step": 472
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.20833333333333337,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1256.0,
+      "completions/mean_length": 909.3333740234375,
+      "completions/mean_terminated_length": 780.2105102539062,
+      "completions/min_length": 455.0,
+      "completions/min_terminated_length": 455.0,
+      "epoch": 67.57142857142857,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 4.351540565490723,
+      "kl": 0.008479526557493955,
+      "learning_rate": 2.332772166583208e-09,
+      "loss": 0.0085,
+      "num_tokens": 12907827.0,
+      "reward": 0.9654167890548706,
+      "reward_std": 1.0205336809158325,
+      "rewards/compute_weighted_reward/mean": 0.965416669845581,
+      "rewards/compute_weighted_reward/std": 1.1710046529769897,
+      "step": 473
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.16666666666666663,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1350.0,
+      "completions/mean_length": 832.5833740234375,
+      "completions/mean_terminated_length": 719.1000366210938,
+      "completions/min_length": 412.0,
+      "completions/min_terminated_length": 412.0,
+      "epoch": 67.71428571428571,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 6.586337089538574,
+      "kl": 0.005711609264835715,
+      "learning_rate": 2.1852399266194312e-09,
+      "loss": 0.0057,
+      "num_tokens": 12934991.0,
+      "reward": 1.3825000524520874,
+      "reward_std": 1.280817985534668,
+      "rewards/compute_weighted_reward/mean": 1.3825000524520874,
+      "rewards/compute_weighted_reward/std": 1.626124382019043,
+      "step": 474
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.25,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1314.0,
+      "completions/mean_length": 806.1666870117188,
+      "completions/mean_terminated_length": 608.2222290039062,
+      "completions/min_length": 155.0,
+      "completions/min_terminated_length": 155.0,
+      "epoch": 67.85714285714286,
+      "frac_reward_zero_std": 0.25,
+      "grad_norm": 2.7453975677490234,
+      "kl": 0.0077794051030650735,
+      "learning_rate": 2.042475040065594e-09,
+      "loss": 0.0078,
+      "num_tokens": 12961863.0,
+      "reward": 1.28000009059906,
+      "reward_std": 0.8226708173751831,
+      "rewards/compute_weighted_reward/mean": 1.28000009059906,
+      "rewards/compute_weighted_reward/std": 1.3074235916137695,
+      "step": 475
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.29166666666666663,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1241.0,
+      "completions/mean_length": 988.3333740234375,
+      "completions/mean_terminated_length": 818.8235473632812,
+      "completions/min_length": 546.0,
+      "completions/min_terminated_length": 546.0,
+      "epoch": 68.0,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 4.332049369812012,
+      "kl": 0.006601755856536329,
+      "learning_rate": 1.9044844650808465e-09,
+      "loss": 0.0066,
+      "num_tokens": 12994193.0,
+      "reward": 0.9479166865348816,
+      "reward_std": 1.4858897924423218,
+      "rewards/compute_weighted_reward/mean": 0.9479167461395264,
+      "rewards/compute_weighted_reward/std": 1.6872307062149048,
+      "step": 476
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.25,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1327.0,
+      "completions/mean_length": 930.75,
+      "completions/mean_terminated_length": 774.3333129882812,
+      "completions/min_length": 366.0,
+      "completions/min_terminated_length": 366.0,
+      "epoch": 68.14285714285714,
+      "frac_reward_zero_std": 0.0,
+      "grad_norm": 5.931741714477539,
+      "kl": 0.007084662560373545,
+      "learning_rate": 1.771274927131139e-09,
+      "loss": 0.0071,
+      "num_tokens": 13023893.0,
+      "reward": 0.3487500548362732,
+      "reward_std": 1.4327483177185059,
+      "rewards/compute_weighted_reward/mean": 0.3487500250339508,
+      "rewards/compute_weighted_reward/std": 1.395942211151123,
+      "step": 477
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.16666666666666663,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1297.0,
+      "completions/mean_length": 822.0833740234375,
+      "completions/mean_terminated_length": 706.5,
+      "completions/min_length": 400.0,
+      "completions/min_terminated_length": 400.0,
+      "epoch": 68.28571428571429,
+      "frac_reward_zero_std": 0.25,
+      "grad_norm": 5.619762420654297,
+      "kl": 0.00922736810753122,
+      "learning_rate": 1.6428529186614193e-09,
+      "loss": 0.0092,
+      "num_tokens": 13051501.0,
+      "reward": 0.8408333659172058,
+      "reward_std": 0.8763567805290222,
+      "rewards/compute_weighted_reward/mean": 0.8408333659172058,
+      "rewards/compute_weighted_reward/std": 1.3136537075042725,
+      "step": 478
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.08333333333333337,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1278.0,
+      "completions/mean_length": 822.0833740234375,
+      "completions/mean_terminated_length": 769.5454711914062,
+      "completions/min_length": 312.0,
+      "completions/min_terminated_length": 312.0,
+      "epoch": 68.42857142857143,
+      "frac_reward_zero_std": 0.25,
+      "grad_norm": 3.2412211894989014,
+      "kl": 0.005566137435380369,
+      "learning_rate": 1.519224698779198e-09,
+      "loss": 0.0056,
+      "num_tokens": 13079013.0,
+      "reward": 1.6216667890548706,
+      "reward_std": 0.771453857421875,
+      "rewards/compute_weighted_reward/mean": 1.6216667890548706,
+      "rewards/compute_weighted_reward/std": 1.126069188117981,
+      "step": 479
+    },
+    {
+      "completion_length": 1400.0,
+      "completions/clipped_ratio": 0.08333333333333337,
+      "completions/max_length": 1400.0,
+      "completions/max_terminated_length": 1385.0,
+      "completions/mean_length": 616.9583740234375,
+      "completions/mean_terminated_length": 545.7727661132812,
+      "completions/min_length": 249.0,
+      "completions/min_terminated_length": 249.0,
+      "epoch": 68.57142857142857,
+      "frac_reward_zero_std": 0.5,
+      "grad_norm": 4.908989429473877,
+      "kl": 0.009014775976538658,
+      "learning_rate": 1.4003962929495127e-09,
+      "loss": 0.009,
+      "num_tokens": 13101392.0,
+      "reward": 1.568333387374878,
+      "reward_std": 0.7887937426567078,
+      "rewards/compute_weighted_reward/mean": 1.5683332681655884,
+      "rewards/compute_weighted_reward/std": 1.244213342666626,
+      "step": 480
     }
   ],
   "logging_steps": 1,
   "max_steps": 500,
-  "num_input_tokens_seen": 12823528,
-  "num_train_epochs": 27,
+  "num_input_tokens_seen": 13101392,
+  "num_train_epochs": 72,
   "save_steps": 10,
   "stateful_callbacks": {
     "TrainerControl": {

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9c7d900446237d1e2d3da3da30c75b8ea971aae913fa39be4480629862dc5a75
+oid sha256:1933debe4019602c345445bbfd993672bc51c69c526e1d1cc5f29a8ef9ce7bbb
 size 7121