Upload folder using huggingface_hub

.gitattributes

@@ -57,3 +57,4 @@ checkpoints-v5.11-b/checkpoint-9216/eval_state.json filter=lfs diff=lfs merge=lf
 checkpoints-v5.11-c/checkpoint-4096/eval_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoints-v5.11-c/checkpoint-10240/eval_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoints-v5.12/checkpoint-11264/eval_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoints-v5.12-b/checkpoint-12288/eval_state.json filter=lfs diff=lfs merge=lfs -text

checkpoints-v5.12-b/checkpoint-12288/ema.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:becb8301744f02b3bee543f98e6c4560263bad84c3059c340c03ab3f90b1188f
+size 55150648

checkpoints-v5.12-b/checkpoint-12288/eval_state.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcd61a1b086998b238df88bfc881eaa8c6b0865b75312290e16a953736bcd25a
+size 60506033

checkpoints-v5.12-b/checkpoint-12288/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:534b3a2eb69f22de17a88cc5758bf6547328c15f2dd780380ca608b9dcc89356
+size 55150680

checkpoints-v5.12-b/checkpoint-12288/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8b8cb3dc1b35065cff4919488935ea316a8ede85090bcdb71cc7a904eb502b0
+size 77725643

checkpoints-v5.12-b/checkpoint-12288/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ecb8eb105f441aa0b9d23a606d4d39b07260a74be26d231f328a9562372100c
+size 14645

checkpoints-v5.12-b/checkpoint-12288/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d75c6d4696a144ff7e32ec64517c03d79b057231c7919118cd52c5ce4205b4e3
+size 1383

checkpoints-v5.12-b/checkpoint-12288/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc7aee33185fd41553948a85c58129c8525a971860371a65bbab2d24b328e431
+size 1465

checkpoints-v5.12-b/checkpoint-12288/trainer_state.json

@@ -0,0 +1,490 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5295752795914409,
+  "eval_steps": 1024,
+  "global_step": 12288,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04413127329928675,
+      "grad_norm": 2.1699578762054443,
+      "learning_rate": 9.990234375e-05,
+      "loss": 14.694396018981934,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04413127329928675,
+      "eval_bleu": 0.05128154542206175,
+      "eval_ce_clean_loss": 3.8575630203238935,
+      "eval_ce_pred_loss": 5.57112058545989,
+      "eval_flow_consistency_loss": 0.21214947685885277,
+      "eval_flow_mse_loss": 0.9729119348627672,
+      "eval_loss": 10.37152462574973,
+      "flow/cos_sim": 0.6056304070741129,
+      "flow/improvement_ratio": 0.9951175209809976,
+      "flow/mag_ratio_mean": 0.596425344821995,
+      "flow/mag_ratio_std": 0.07446127475451814,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04413127329928675,
+      "eval_bleu": 0.05128154542206175,
+      "eval_ce_clean_loss": 3.8575630203238935,
+      "eval_ce_pred_loss": 5.57112058545989,
+      "eval_flow_consistency_loss": 0.21214947685885277,
+      "eval_flow_mse_loss": 0.9729119348627672,
+      "eval_loss": 10.37152462574973,
+      "eval_runtime": 210.6333,
+      "eval_samples_per_second": 142.428,
+      "eval_steps_per_second": 2.227,
+      "flow/cos_sim": 0.6056304070741129,
+      "flow/improvement_ratio": 0.9951175209809976,
+      "flow/mag_ratio_mean": 0.596425344821995,
+      "flow/mag_ratio_std": 0.07446127475451814,
+      "step": 1024
+    },
+    {
+      "epoch": 0.0882625465985735,
+      "grad_norm": 2.5218427181243896,
+      "learning_rate": 9.9476028157316e-05,
+      "loss": 8.70075798034668,
+      "step": 2048
+    },
+    {
+      "epoch": 0.0882625465985735,
+      "eval_bleu": 0.23857394758705833,
+      "eval_ce_clean_loss": 1.3269661532790422,
+      "eval_ce_pred_loss": 3.7169905786575286,
+      "eval_flow_consistency_loss": 0.17732289093516784,
+      "eval_flow_mse_loss": 1.097706873788,
+      "eval_loss": 7.496601121003694,
+      "flow/cos_sim": 0.5965837963354359,
+      "flow/improvement_ratio": 0.9938774880315704,
+      "flow/mag_ratio_mean": 0.5317577726042855,
+      "flow/mag_ratio_std": 0.06900304873614932,
+      "step": 2048
+    },
+    {
+      "epoch": 0.0882625465985735,
+      "eval_bleu": 0.23857394758705833,
+      "eval_ce_clean_loss": 1.3269661532790422,
+      "eval_ce_pred_loss": 3.7169905786575286,
+      "eval_flow_consistency_loss": 0.17732289093516784,
+      "eval_flow_mse_loss": 1.097706873788,
+      "eval_loss": 7.496601121003694,
+      "eval_runtime": 208.2982,
+      "eval_samples_per_second": 144.024,
+      "eval_steps_per_second": 2.252,
+      "flow/cos_sim": 0.5965837963354359,
+      "flow/improvement_ratio": 0.9938774880315704,
+      "flow/mag_ratio_mean": 0.5317577726042855,
+      "flow/mag_ratio_std": 0.06900304873614932,
+      "step": 2048
+    },
+    {
+      "epoch": 0.13239381989786023,
+      "grad_norm": 2.14751935005188,
+      "learning_rate": 9.791307026072513e-05,
+      "loss": 6.987759113311768,
+      "step": 3072
+    },
+    {
+      "epoch": 0.13239381989786023,
+      "eval_bleu": 0.3493089368796349,
+      "eval_ce_clean_loss": 0.5745006232881851,
+      "eval_ce_pred_loss": 3.0287138033014878,
+      "eval_flow_consistency_loss": 0.18727965287562373,
+      "eval_flow_mse_loss": 1.1055078414965793,
+      "eval_loss": 6.408606252690622,
+      "flow/cos_sim": 0.6098656960641906,
+      "flow/improvement_ratio": 0.9930039020235351,
+      "flow/mag_ratio_mean": 0.554409195238085,
+      "flow/mag_ratio_std": 0.08226730507701191,
+      "step": 3072
+    },
+    {
+      "epoch": 0.13239381989786023,
+      "eval_bleu": 0.3493089368796349,
+      "eval_ce_clean_loss": 0.5745006232881851,
+      "eval_ce_pred_loss": 3.0287138033014878,
+      "eval_flow_consistency_loss": 0.18727965287562373,
+      "eval_flow_mse_loss": 1.1055078414965793,
+      "eval_loss": 6.408606252690622,
+      "eval_runtime": 207.7037,
+      "eval_samples_per_second": 144.437,
+      "eval_steps_per_second": 2.258,
+      "flow/cos_sim": 0.6098656960641906,
+      "flow/improvement_ratio": 0.9930039020235351,
+      "flow/mag_ratio_mean": 0.554409195238085,
+      "flow/mag_ratio_std": 0.08226730507701191,
+      "step": 3072
+    },
+    {
+      "epoch": 0.176525093197147,
+      "grad_norm": 1.9934877157211304,
+      "learning_rate": 9.534693146185996e-05,
+      "loss": 6.277302265167236,
+      "step": 4096
+    },
+    {
+      "epoch": 0.176525093197147,
+      "eval_bleu": 0.4002615891374953,
+      "eval_ce_clean_loss": 0.30466608197958484,
+      "eval_ce_pred_loss": 2.713986599114912,
+      "eval_flow_consistency_loss": 0.19233520605416696,
+      "eval_flow_mse_loss": 1.1293878446001488,
+      "eval_loss": 5.949396448603062,
+      "flow/cos_sim": 0.6327927858590572,
+      "flow/improvement_ratio": 0.9936480792854895,
+      "flow/mag_ratio_mean": 0.5879036461366519,
+      "flow/mag_ratio_std": 0.10289596083131172,
+      "step": 4096
+    },
+    {
+      "epoch": 0.176525093197147,
+      "eval_bleu": 0.4002615891374953,
+      "eval_ce_clean_loss": 0.30466608197958484,
+      "eval_ce_pred_loss": 2.713986599114912,
+      "eval_flow_consistency_loss": 0.19233520605416696,
+      "eval_flow_mse_loss": 1.1293878446001488,
+      "eval_loss": 5.949396448603062,
+      "eval_runtime": 207.8349,
+      "eval_samples_per_second": 144.345,
+      "eval_steps_per_second": 2.257,
+      "flow/cos_sim": 0.6327927858590572,
+      "flow/improvement_ratio": 0.9936480792854895,
+      "flow/mag_ratio_mean": 0.5879036461366519,
+      "flow/mag_ratio_std": 0.10289596083131172,
+      "step": 4096
+    },
+    {
+      "epoch": 0.22065636649643372,
+      "grad_norm": 2.809199810028076,
+      "learning_rate": 9.18264920723673e-05,
+      "loss": 5.952817440032959,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22065636649643372,
+      "eval_bleu": 0.42835581482136026,
+      "eval_ce_clean_loss": 0.1883118874482763,
+      "eval_ce_pred_loss": 2.513261323798694,
+      "eval_flow_consistency_loss": 0.2029014579268661,
+      "eval_flow_mse_loss": 1.1447690499108483,
+      "eval_loss": 5.668197899230762,
+      "flow/cos_sim": 0.6509825842721122,
+      "flow/improvement_ratio": 0.9934964183805339,
+      "flow/mag_ratio_mean": 0.6140208387934069,
+      "flow/mag_ratio_std": 0.12194133851764553,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22065636649643372,
+      "eval_bleu": 0.42835581482136026,
+      "eval_ce_clean_loss": 0.1883118874482763,
+      "eval_ce_pred_loss": 2.513261323798694,
+      "eval_flow_consistency_loss": 0.2029014579268661,
+      "eval_flow_mse_loss": 1.1447690499108483,
+      "eval_loss": 5.668197899230762,
+      "eval_runtime": 209.54,
+      "eval_samples_per_second": 143.171,
+      "eval_steps_per_second": 2.238,
+      "flow/cos_sim": 0.6509825842721122,
+      "flow/improvement_ratio": 0.9934964183805339,
+      "flow/mag_ratio_mean": 0.6140208387934069,
+      "flow/mag_ratio_std": 0.12194133851764553,
+      "step": 5120
+    },
+    {
+      "epoch": 0.26478763979572045,
+      "grad_norm": 2.2502171993255615,
+      "learning_rate": 8.74324003722993e-05,
+      "loss": 5.733938217163086,
+      "step": 6144
+    },
+    {
+      "epoch": 0.26478763979572045,
+      "eval_bleu": 0.45224998228939733,
+      "eval_ce_clean_loss": 0.12550937678259827,
+      "eval_ce_pred_loss": 2.4016613040143238,
+      "eval_flow_consistency_loss": 0.20323555880009747,
+      "eval_flow_mse_loss": 1.1755171974838923,
+      "eval_loss": 5.540907144546509,
+      "flow/cos_sim": 0.6656121532800101,
+      "flow/improvement_ratio": 0.993739642822412,
+      "flow/mag_ratio_mean": 0.6285864149075328,
+      "flow/mag_ratio_std": 0.14186543162697668,
+      "step": 6144
+    },
+    {
+      "epoch": 0.26478763979572045,
+      "eval_bleu": 0.45224998228939733,
+      "eval_ce_clean_loss": 0.12550937678259827,
+      "eval_ce_pred_loss": 2.4016613040143238,
+      "eval_flow_consistency_loss": 0.20323555880009747,
+      "eval_flow_mse_loss": 1.1755171974838923,
+      "eval_loss": 5.540907144546509,
+      "eval_runtime": 210.3673,
+      "eval_samples_per_second": 142.608,
+      "eval_steps_per_second": 2.229,
+      "flow/cos_sim": 0.6656121532800101,
+      "flow/improvement_ratio": 0.993739642822412,
+      "flow/mag_ratio_mean": 0.6285864149075328,
+      "flow/mag_ratio_std": 0.14186543162697668,
+      "step": 6144
+    },
+    {
+      "epoch": 0.30891891309500724,
+      "grad_norm": 1.2597520351409912,
+      "learning_rate": 8.22483558761947e-05,
+      "loss": 5.561953544616699,
+      "step": 7168
+    },
+    {
+      "epoch": 0.30891891309500724,
+      "eval_bleu": 0.4687065651483432,
+      "eval_ce_clean_loss": 0.08836616801300537,
+      "eval_ce_pred_loss": 2.2620769802695397,
+      "eval_flow_consistency_loss": 0.2063061216238465,
+      "eval_flow_mse_loss": 1.1843450824334931,
+      "eval_loss": 5.343514618843095,
+      "flow/cos_sim": 0.6833345525300325,
+      "flow/improvement_ratio": 0.9944396457438276,
+      "flow/mag_ratio_mean": 0.6510747799487002,
+      "flow/mag_ratio_std": 0.15373703074861944,
+      "step": 7168
+    },
+    {
+      "epoch": 0.30891891309500724,
+      "eval_bleu": 0.4687065651483432,
+      "eval_ce_clean_loss": 0.08836616801300537,
+      "eval_ce_pred_loss": 2.2620769802695397,
+      "eval_flow_consistency_loss": 0.2063061216238465,
+      "eval_flow_mse_loss": 1.1843450824334931,
+      "eval_loss": 5.343514618843095,
+      "eval_runtime": 212.5372,
+      "eval_samples_per_second": 141.152,
+      "eval_steps_per_second": 2.207,
+      "flow/cos_sim": 0.6833345525300325,
+      "flow/improvement_ratio": 0.9944396457438276,
+      "flow/mag_ratio_mean": 0.6510747799487002,
+      "flow/mag_ratio_std": 0.15373703074861944,
+      "step": 7168
+    },
+    {
+      "epoch": 0.353050186394294,
+      "grad_norm": 2.1638965606689453,
+      "learning_rate": 7.639311770076283e-05,
+      "loss": 5.4200921058654785,
+      "step": 8192
+    },
+    {
+      "epoch": 0.353050186394294,
+      "eval_bleu": 0.4806818081298782,
+      "eval_ce_clean_loss": 0.06611348781536129,
+      "eval_ce_pred_loss": 2.205830120582825,
+      "eval_flow_consistency_loss": 0.21406505789075578,
+      "eval_flow_mse_loss": 1.2107092700024913,
+      "eval_loss": 5.302672486315404,
+      "flow/cos_sim": 0.6940237746309879,
+      "flow/improvement_ratio": 0.993346476732795,
+      "flow/mag_ratio_mean": 0.6572817430567386,
+      "flow/mag_ratio_std": 0.1605924148676492,
+      "step": 8192
+    },
+    {
+      "epoch": 0.353050186394294,
+      "eval_bleu": 0.4806818081298782,
+      "eval_ce_clean_loss": 0.06611348781536129,
+      "eval_ce_pred_loss": 2.205830120582825,
+      "eval_flow_consistency_loss": 0.21406505789075578,
+      "eval_flow_mse_loss": 1.2107092700024913,
+      "eval_loss": 5.302672486315404,
+      "eval_runtime": 210.6328,
+      "eval_samples_per_second": 142.428,
+      "eval_steps_per_second": 2.227,
+      "flow/cos_sim": 0.6940237746309879,
+      "flow/improvement_ratio": 0.993346476732795,
+      "flow/mag_ratio_mean": 0.6572817430567386,
+      "flow/mag_ratio_std": 0.1605924148676492,
+      "step": 8192
+    },
+    {
+      "epoch": 0.3971814596935807,
+      "grad_norm": 3.1213200092315674,
+      "learning_rate": 6.998470950469718e-05,
+      "loss": 5.320468425750732,
+      "step": 9216
+    },
+    {
+      "epoch": 0.3971814596935807,
+      "eval_bleu": 0.49321486760812644,
+      "eval_ce_clean_loss": 0.051587040851841855,
+      "eval_ce_pred_loss": 2.1326764727960517,
+      "eval_flow_consistency_loss": 0.20684453628973157,
+      "eval_flow_mse_loss": 1.2189766706179963,
+      "eval_loss": 5.198114048951725,
+      "flow/cos_sim": 0.7069450036040755,
+      "flow/improvement_ratio": 0.9930796729984568,
+      "flow/mag_ratio_mean": 0.6787635914044086,
+      "flow/mag_ratio_std": 0.16792897991280056,
+      "step": 9216
+    },
+    {
+      "epoch": 0.3971814596935807,
+      "eval_bleu": 0.49321486760812644,
+      "eval_ce_clean_loss": 0.051587040851841855,
+      "eval_ce_pred_loss": 2.1326764727960517,
+      "eval_flow_consistency_loss": 0.20684453628973157,
+      "eval_flow_mse_loss": 1.2189766706179963,
+      "eval_loss": 5.198114048951725,
+      "eval_runtime": 210.2128,
+      "eval_samples_per_second": 142.713,
+      "eval_steps_per_second": 2.231,
+      "flow/cos_sim": 0.7069450036040755,
+      "flow/improvement_ratio": 0.9930796729984568,
+      "flow/mag_ratio_mean": 0.6787635914044086,
+      "flow/mag_ratio_std": 0.16792897991280056,
+      "step": 9216
+    },
+    {
+      "epoch": 0.44131273299286744,
+      "grad_norm": 1.7769527435302734,
+      "learning_rate": 6.315061173955019e-05,
+      "loss": 5.260382652282715,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44131273299286744,
+      "eval_bleu": 0.5003817272012132,
+      "eval_ce_clean_loss": 0.04119707541520407,
+      "eval_ce_pred_loss": 2.080879623447654,
+      "eval_flow_consistency_loss": 0.20481373865340055,
+      "eval_flow_mse_loss": 1.2196696361244868,
+      "eval_loss": 5.118794566786874,
+      "flow/cos_sim": 0.7191916523711768,
+      "flow/improvement_ratio": 0.993994485340647,
+      "flow/mag_ratio_mean": 0.6901472842515405,
+      "flow/mag_ratio_std": 0.17046165418650297,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44131273299286744,
+      "eval_bleu": 0.5003817272012132,
+      "eval_ce_clean_loss": 0.04119707541520407,
+      "eval_ce_pred_loss": 2.080879623447654,
+      "eval_flow_consistency_loss": 0.20481373865340055,
+      "eval_flow_mse_loss": 1.2196696361244868,
+      "eval_loss": 5.118794566786874,
+      "eval_runtime": 209.7844,
+      "eval_samples_per_second": 143.004,
+      "eval_steps_per_second": 2.236,
+      "flow/cos_sim": 0.7191916523711768,
+      "flow/improvement_ratio": 0.993994485340647,
+      "flow/mag_ratio_mean": 0.6901472842515405,
+      "flow/mag_ratio_std": 0.17046165418650297,
+      "step": 10240
+    },
+    {
+      "epoch": 0.4854440062921542,
+      "grad_norm": 2.305457592010498,
+      "learning_rate": 5.604035379537632e-05,
+      "loss": 5.207396030426025,
+      "step": 11264
+    },
+    {
+      "epoch": 0.4854440062921542,
+      "eval_bleu": 0.5041758930303886,
+      "eval_ce_clean_loss": 0.03406135918203193,
+      "eval_ce_pred_loss": 2.032164579007163,
+      "eval_flow_consistency_loss": 0.1978874005107229,
+      "eval_flow_mse_loss": 1.2266178403073535,
+      "eval_loss": 5.049889673302168,
+      "flow/cos_sim": 0.7285718597582916,
+      "flow/improvement_ratio": 0.995131308081816,
+      "flow/mag_ratio_mean": 0.7023200071188432,
+      "flow/mag_ratio_std": 0.1736910315528353,
+      "step": 11264
+    },
+    {
+      "epoch": 0.4854440062921542,
+      "eval_bleu": 0.5041758930303886,
+      "eval_ce_clean_loss": 0.03406135918203193,
+      "eval_ce_pred_loss": 2.032164579007163,
+      "eval_flow_consistency_loss": 0.1978874005107229,
+      "eval_flow_mse_loss": 1.2266178403073535,
+      "eval_loss": 5.049889673302168,
+      "eval_runtime": 208.1179,
+      "eval_samples_per_second": 144.149,
+      "eval_steps_per_second": 2.254,
+      "flow/cos_sim": 0.7285718597582916,
+      "flow/improvement_ratio": 0.995131308081816,
+      "flow/mag_ratio_mean": 0.7023200071188432,
+      "flow/mag_ratio_std": 0.1736910315528353,
+      "step": 11264
+    },
+    {
+      "epoch": 0.5295752795914409,
+      "grad_norm": 1.614006757736206,
+      "learning_rate": 4.881032966918056e-05,
+      "loss": 5.130342483520508,
+      "step": 12288
+    },
+    {
+      "epoch": 0.5295752795914409,
+      "eval_bleu": 0.514033522638653,
+      "eval_ce_clean_loss": 0.028551021900607834,
+      "eval_ce_pred_loss": 2.0023169781861783,
+      "eval_flow_consistency_loss": 0.1769014176592898,
+      "eval_flow_mse_loss": 1.2329019299193995,
+      "eval_loss": 4.997204606212787,
+      "flow/cos_sim": 0.737011756088688,
+      "flow/improvement_ratio": 0.9937192791306388,
+      "flow/mag_ratio_mean": 0.707237169813754,
+      "flow/mag_ratio_std": 0.17565249913791095,
+      "step": 12288
+    },
+    {
+      "epoch": 0.5295752795914409,
+      "eval_bleu": 0.514033522638653,
+      "eval_ce_clean_loss": 0.028551021900607834,
+      "eval_ce_pred_loss": 2.0023169781861783,
+      "eval_flow_consistency_loss": 0.1769014176592898,
+      "eval_flow_mse_loss": 1.2329019299193995,
+      "eval_loss": 4.997204606212787,
+      "eval_runtime": 209.7397,
+      "eval_samples_per_second": 143.034,
+      "eval_steps_per_second": 2.236,
+      "flow/cos_sim": 0.737011756088688,
+      "flow/improvement_ratio": 0.9937192791306388,
+      "flow/mag_ratio_mean": 0.707237169813754,
+      "flow/mag_ratio_std": 0.17565249913791095,
+      "step": 12288
+    }
+  ],
+  "logging_steps": 1024,
+  "max_steps": 23204,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1024,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoints-v5.12-b/checkpoint-12288/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8469bbc81a2ba0be2f5b44007faafd15c75615abe30f4f4e56171816d31caa5b
+size 5137