abdiharyadi
commited on
Commit
•
faf95da
1
Parent(s):
4bf1f7f
Training in progress, step 10800, checkpoint
Browse files
last-checkpoint/model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1575259780
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7e207271ed9fc29b099863defab32d4952b4913bf12a40b3e58c16c30ace3354
|
3 |
size 1575259780
|
last-checkpoint/optimizer.pt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 3150397656
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fef1272e9932d85f4a0434d895c0520855facd99c96206dd3c6f6de6dc78ba1a
|
3 |
size 3150397656
|
last-checkpoint/rng_state.pth
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 14244
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:725551c540c5935ca09ad257e1ec602a9cde53567df7545ad488987db5f81d2a
|
3 |
size 14244
|
last-checkpoint/scheduler.pt
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1064
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f7a14dd5679ca365b7036c52129c5a5a2c8f2459be0b4d29ecb21c4b219dee84
|
3 |
size 1064
|
last-checkpoint/trainer_state.json
CHANGED
@@ -1,9 +1,9 @@
|
|
1 |
{
|
2 |
-
"best_metric":
|
3 |
-
"best_model_checkpoint": "/kaggle/working/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted-amr-generation-v2-fted/checkpoint-
|
4 |
-
"epoch":
|
5 |
"eval_steps": 3600,
|
6 |
-
"global_step":
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
@@ -2193,6 +2193,1096 @@
|
|
2193 |
"eval_samples_per_second": 8.751,
|
2194 |
"eval_steps_per_second": 1.75,
|
2195 |
"step": 7200
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2196 |
}
|
2197 |
],
|
2198 |
"logging_steps": 20,
|
@@ -2212,7 +3302,7 @@
|
|
2212 |
"attributes": {}
|
2213 |
}
|
2214 |
},
|
2215 |
-
"total_flos":
|
2216 |
"train_batch_size": 5,
|
2217 |
"trial_name": null,
|
2218 |
"trial_params": null
|
|
|
1 |
{
|
2 |
+
"best_metric": 49.121,
|
3 |
+
"best_model_checkpoint": "/kaggle/working/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted-amr-generation-v2-fted/checkpoint-10800",
|
4 |
+
"epoch": 119.20529801324503,
|
5 |
"eval_steps": 3600,
|
6 |
+
"global_step": 10800,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
|
|
2193 |
"eval_samples_per_second": 8.751,
|
2194 |
"eval_steps_per_second": 1.75,
|
2195 |
"step": 7200
|
2196 |
+
},
|
2197 |
+
{
|
2198 |
+
"epoch": 79.69094922737307,
|
2199 |
+
"learning_rate": 8.899303135888502e-07,
|
2200 |
+
"loss": 1.4294,
|
2201 |
+
"step": 7220
|
2202 |
+
},
|
2203 |
+
{
|
2204 |
+
"epoch": 79.91169977924945,
|
2205 |
+
"learning_rate": 8.896167247386759e-07,
|
2206 |
+
"loss": 1.4691,
|
2207 |
+
"step": 7240
|
2208 |
+
},
|
2209 |
+
{
|
2210 |
+
"epoch": 80.13245033112582,
|
2211 |
+
"learning_rate": 8.893031358885017e-07,
|
2212 |
+
"loss": 1.4619,
|
2213 |
+
"step": 7260
|
2214 |
+
},
|
2215 |
+
{
|
2216 |
+
"epoch": 80.3532008830022,
|
2217 |
+
"learning_rate": 8.889895470383275e-07,
|
2218 |
+
"loss": 1.4952,
|
2219 |
+
"step": 7280
|
2220 |
+
},
|
2221 |
+
{
|
2222 |
+
"epoch": 80.57395143487858,
|
2223 |
+
"learning_rate": 8.886759581881533e-07,
|
2224 |
+
"loss": 1.4284,
|
2225 |
+
"step": 7300
|
2226 |
+
},
|
2227 |
+
{
|
2228 |
+
"epoch": 80.79470198675497,
|
2229 |
+
"learning_rate": 8.88362369337979e-07,
|
2230 |
+
"loss": 1.3992,
|
2231 |
+
"step": 7320
|
2232 |
+
},
|
2233 |
+
{
|
2234 |
+
"epoch": 81.01545253863135,
|
2235 |
+
"learning_rate": 8.880487804878048e-07,
|
2236 |
+
"loss": 1.4512,
|
2237 |
+
"step": 7340
|
2238 |
+
},
|
2239 |
+
{
|
2240 |
+
"epoch": 81.23620309050773,
|
2241 |
+
"learning_rate": 8.877351916376307e-07,
|
2242 |
+
"loss": 1.4298,
|
2243 |
+
"step": 7360
|
2244 |
+
},
|
2245 |
+
{
|
2246 |
+
"epoch": 81.45695364238411,
|
2247 |
+
"learning_rate": 8.874216027874564e-07,
|
2248 |
+
"loss": 1.4442,
|
2249 |
+
"step": 7380
|
2250 |
+
},
|
2251 |
+
{
|
2252 |
+
"epoch": 81.67770419426049,
|
2253 |
+
"learning_rate": 8.871080139372822e-07,
|
2254 |
+
"loss": 1.4112,
|
2255 |
+
"step": 7400
|
2256 |
+
},
|
2257 |
+
{
|
2258 |
+
"epoch": 81.89845474613686,
|
2259 |
+
"learning_rate": 8.867944250871079e-07,
|
2260 |
+
"loss": 1.4365,
|
2261 |
+
"step": 7420
|
2262 |
+
},
|
2263 |
+
{
|
2264 |
+
"epoch": 82.11920529801324,
|
2265 |
+
"learning_rate": 8.864808362369338e-07,
|
2266 |
+
"loss": 1.5076,
|
2267 |
+
"step": 7440
|
2268 |
+
},
|
2269 |
+
{
|
2270 |
+
"epoch": 82.33995584988962,
|
2271 |
+
"learning_rate": 8.861672473867594e-07,
|
2272 |
+
"loss": 1.4004,
|
2273 |
+
"step": 7460
|
2274 |
+
},
|
2275 |
+
{
|
2276 |
+
"epoch": 82.560706401766,
|
2277 |
+
"learning_rate": 8.858536585365853e-07,
|
2278 |
+
"loss": 1.4641,
|
2279 |
+
"step": 7480
|
2280 |
+
},
|
2281 |
+
{
|
2282 |
+
"epoch": 82.78145695364239,
|
2283 |
+
"learning_rate": 8.855400696864112e-07,
|
2284 |
+
"loss": 1.4321,
|
2285 |
+
"step": 7500
|
2286 |
+
},
|
2287 |
+
{
|
2288 |
+
"epoch": 83.00220750551877,
|
2289 |
+
"learning_rate": 8.85226480836237e-07,
|
2290 |
+
"loss": 1.4592,
|
2291 |
+
"step": 7520
|
2292 |
+
},
|
2293 |
+
{
|
2294 |
+
"epoch": 83.22295805739515,
|
2295 |
+
"learning_rate": 8.849128919860627e-07,
|
2296 |
+
"loss": 1.4101,
|
2297 |
+
"step": 7540
|
2298 |
+
},
|
2299 |
+
{
|
2300 |
+
"epoch": 83.44370860927152,
|
2301 |
+
"learning_rate": 8.845993031358884e-07,
|
2302 |
+
"loss": 1.4455,
|
2303 |
+
"step": 7560
|
2304 |
+
},
|
2305 |
+
{
|
2306 |
+
"epoch": 83.6644591611479,
|
2307 |
+
"learning_rate": 8.842857142857143e-07,
|
2308 |
+
"loss": 1.4175,
|
2309 |
+
"step": 7580
|
2310 |
+
},
|
2311 |
+
{
|
2312 |
+
"epoch": 83.88520971302428,
|
2313 |
+
"learning_rate": 8.8397212543554e-07,
|
2314 |
+
"loss": 1.4389,
|
2315 |
+
"step": 7600
|
2316 |
+
},
|
2317 |
+
{
|
2318 |
+
"epoch": 84.10596026490066,
|
2319 |
+
"learning_rate": 8.836585365853658e-07,
|
2320 |
+
"loss": 1.4963,
|
2321 |
+
"step": 7620
|
2322 |
+
},
|
2323 |
+
{
|
2324 |
+
"epoch": 84.32671081677704,
|
2325 |
+
"learning_rate": 8.833449477351916e-07,
|
2326 |
+
"loss": 1.4683,
|
2327 |
+
"step": 7640
|
2328 |
+
},
|
2329 |
+
{
|
2330 |
+
"epoch": 84.54746136865342,
|
2331 |
+
"learning_rate": 8.830313588850174e-07,
|
2332 |
+
"loss": 1.4528,
|
2333 |
+
"step": 7660
|
2334 |
+
},
|
2335 |
+
{
|
2336 |
+
"epoch": 84.76821192052981,
|
2337 |
+
"learning_rate": 8.827177700348431e-07,
|
2338 |
+
"loss": 1.4128,
|
2339 |
+
"step": 7680
|
2340 |
+
},
|
2341 |
+
{
|
2342 |
+
"epoch": 84.98896247240619,
|
2343 |
+
"learning_rate": 8.824041811846689e-07,
|
2344 |
+
"loss": 1.4012,
|
2345 |
+
"step": 7700
|
2346 |
+
},
|
2347 |
+
{
|
2348 |
+
"epoch": 85.20971302428256,
|
2349 |
+
"learning_rate": 8.820905923344947e-07,
|
2350 |
+
"loss": 1.4448,
|
2351 |
+
"step": 7720
|
2352 |
+
},
|
2353 |
+
{
|
2354 |
+
"epoch": 85.43046357615894,
|
2355 |
+
"learning_rate": 8.817770034843205e-07,
|
2356 |
+
"loss": 1.3746,
|
2357 |
+
"step": 7740
|
2358 |
+
},
|
2359 |
+
{
|
2360 |
+
"epoch": 85.65121412803532,
|
2361 |
+
"learning_rate": 8.814634146341464e-07,
|
2362 |
+
"loss": 1.4086,
|
2363 |
+
"step": 7760
|
2364 |
+
},
|
2365 |
+
{
|
2366 |
+
"epoch": 85.8719646799117,
|
2367 |
+
"learning_rate": 8.811498257839721e-07,
|
2368 |
+
"loss": 1.4774,
|
2369 |
+
"step": 7780
|
2370 |
+
},
|
2371 |
+
{
|
2372 |
+
"epoch": 86.09271523178808,
|
2373 |
+
"learning_rate": 8.808362369337979e-07,
|
2374 |
+
"loss": 1.4534,
|
2375 |
+
"step": 7800
|
2376 |
+
},
|
2377 |
+
{
|
2378 |
+
"epoch": 86.31346578366445,
|
2379 |
+
"learning_rate": 8.805226480836237e-07,
|
2380 |
+
"loss": 1.3844,
|
2381 |
+
"step": 7820
|
2382 |
+
},
|
2383 |
+
{
|
2384 |
+
"epoch": 86.53421633554083,
|
2385 |
+
"learning_rate": 8.802090592334494e-07,
|
2386 |
+
"loss": 1.4751,
|
2387 |
+
"step": 7840
|
2388 |
+
},
|
2389 |
+
{
|
2390 |
+
"epoch": 86.75496688741723,
|
2391 |
+
"learning_rate": 8.798954703832752e-07,
|
2392 |
+
"loss": 1.4002,
|
2393 |
+
"step": 7860
|
2394 |
+
},
|
2395 |
+
{
|
2396 |
+
"epoch": 86.9757174392936,
|
2397 |
+
"learning_rate": 8.79581881533101e-07,
|
2398 |
+
"loss": 1.4659,
|
2399 |
+
"step": 7880
|
2400 |
+
},
|
2401 |
+
{
|
2402 |
+
"epoch": 87.19646799116998,
|
2403 |
+
"learning_rate": 8.792682926829268e-07,
|
2404 |
+
"loss": 1.4322,
|
2405 |
+
"step": 7900
|
2406 |
+
},
|
2407 |
+
{
|
2408 |
+
"epoch": 87.41721854304636,
|
2409 |
+
"learning_rate": 8.789547038327526e-07,
|
2410 |
+
"loss": 1.4278,
|
2411 |
+
"step": 7920
|
2412 |
+
},
|
2413 |
+
{
|
2414 |
+
"epoch": 87.63796909492274,
|
2415 |
+
"learning_rate": 8.786411149825783e-07,
|
2416 |
+
"loss": 1.453,
|
2417 |
+
"step": 7940
|
2418 |
+
},
|
2419 |
+
{
|
2420 |
+
"epoch": 87.85871964679912,
|
2421 |
+
"learning_rate": 8.783275261324042e-07,
|
2422 |
+
"loss": 1.4371,
|
2423 |
+
"step": 7960
|
2424 |
+
},
|
2425 |
+
{
|
2426 |
+
"epoch": 88.0794701986755,
|
2427 |
+
"learning_rate": 8.780139372822298e-07,
|
2428 |
+
"loss": 1.4775,
|
2429 |
+
"step": 7980
|
2430 |
+
},
|
2431 |
+
{
|
2432 |
+
"epoch": 88.30022075055187,
|
2433 |
+
"learning_rate": 8.777003484320557e-07,
|
2434 |
+
"loss": 1.4058,
|
2435 |
+
"step": 8000
|
2436 |
+
},
|
2437 |
+
{
|
2438 |
+
"epoch": 88.52097130242825,
|
2439 |
+
"learning_rate": 8.773867595818815e-07,
|
2440 |
+
"loss": 1.4119,
|
2441 |
+
"step": 8020
|
2442 |
+
},
|
2443 |
+
{
|
2444 |
+
"epoch": 88.74172185430463,
|
2445 |
+
"learning_rate": 8.770731707317073e-07,
|
2446 |
+
"loss": 1.4316,
|
2447 |
+
"step": 8040
|
2448 |
+
},
|
2449 |
+
{
|
2450 |
+
"epoch": 88.96247240618102,
|
2451 |
+
"learning_rate": 8.767595818815331e-07,
|
2452 |
+
"loss": 1.4198,
|
2453 |
+
"step": 8060
|
2454 |
+
},
|
2455 |
+
{
|
2456 |
+
"epoch": 89.1832229580574,
|
2457 |
+
"learning_rate": 8.764459930313588e-07,
|
2458 |
+
"loss": 1.4434,
|
2459 |
+
"step": 8080
|
2460 |
+
},
|
2461 |
+
{
|
2462 |
+
"epoch": 89.40397350993378,
|
2463 |
+
"learning_rate": 8.761324041811848e-07,
|
2464 |
+
"loss": 1.4309,
|
2465 |
+
"step": 8100
|
2466 |
+
},
|
2467 |
+
{
|
2468 |
+
"epoch": 89.62472406181016,
|
2469 |
+
"learning_rate": 8.758188153310104e-07,
|
2470 |
+
"loss": 1.3757,
|
2471 |
+
"step": 8120
|
2472 |
+
},
|
2473 |
+
{
|
2474 |
+
"epoch": 89.84547461368653,
|
2475 |
+
"learning_rate": 8.755052264808362e-07,
|
2476 |
+
"loss": 1.479,
|
2477 |
+
"step": 8140
|
2478 |
+
},
|
2479 |
+
{
|
2480 |
+
"epoch": 90.06622516556291,
|
2481 |
+
"learning_rate": 8.751916376306619e-07,
|
2482 |
+
"loss": 1.3926,
|
2483 |
+
"step": 8160
|
2484 |
+
},
|
2485 |
+
{
|
2486 |
+
"epoch": 90.28697571743929,
|
2487 |
+
"learning_rate": 8.748780487804878e-07,
|
2488 |
+
"loss": 1.4629,
|
2489 |
+
"step": 8180
|
2490 |
+
},
|
2491 |
+
{
|
2492 |
+
"epoch": 90.50772626931567,
|
2493 |
+
"learning_rate": 8.745644599303135e-07,
|
2494 |
+
"loss": 1.3793,
|
2495 |
+
"step": 8200
|
2496 |
+
},
|
2497 |
+
{
|
2498 |
+
"epoch": 90.72847682119205,
|
2499 |
+
"learning_rate": 8.742508710801393e-07,
|
2500 |
+
"loss": 1.4113,
|
2501 |
+
"step": 8220
|
2502 |
+
},
|
2503 |
+
{
|
2504 |
+
"epoch": 90.94922737306844,
|
2505 |
+
"learning_rate": 8.739372822299651e-07,
|
2506 |
+
"loss": 1.4187,
|
2507 |
+
"step": 8240
|
2508 |
+
},
|
2509 |
+
{
|
2510 |
+
"epoch": 91.16997792494482,
|
2511 |
+
"learning_rate": 8.73623693379791e-07,
|
2512 |
+
"loss": 1.4466,
|
2513 |
+
"step": 8260
|
2514 |
+
},
|
2515 |
+
{
|
2516 |
+
"epoch": 91.3907284768212,
|
2517 |
+
"learning_rate": 8.733101045296167e-07,
|
2518 |
+
"loss": 1.4219,
|
2519 |
+
"step": 8280
|
2520 |
+
},
|
2521 |
+
{
|
2522 |
+
"epoch": 91.61147902869757,
|
2523 |
+
"learning_rate": 8.729965156794424e-07,
|
2524 |
+
"loss": 1.4783,
|
2525 |
+
"step": 8300
|
2526 |
+
},
|
2527 |
+
{
|
2528 |
+
"epoch": 91.83222958057395,
|
2529 |
+
"learning_rate": 8.726829268292683e-07,
|
2530 |
+
"loss": 1.4111,
|
2531 |
+
"step": 8320
|
2532 |
+
},
|
2533 |
+
{
|
2534 |
+
"epoch": 92.05298013245033,
|
2535 |
+
"learning_rate": 8.723693379790941e-07,
|
2536 |
+
"loss": 1.427,
|
2537 |
+
"step": 8340
|
2538 |
+
},
|
2539 |
+
{
|
2540 |
+
"epoch": 92.27373068432671,
|
2541 |
+
"learning_rate": 8.720557491289198e-07,
|
2542 |
+
"loss": 1.414,
|
2543 |
+
"step": 8360
|
2544 |
+
},
|
2545 |
+
{
|
2546 |
+
"epoch": 92.49448123620309,
|
2547 |
+
"learning_rate": 8.717421602787456e-07,
|
2548 |
+
"loss": 1.4097,
|
2549 |
+
"step": 8380
|
2550 |
+
},
|
2551 |
+
{
|
2552 |
+
"epoch": 92.71523178807946,
|
2553 |
+
"learning_rate": 8.714285714285715e-07,
|
2554 |
+
"loss": 1.4818,
|
2555 |
+
"step": 8400
|
2556 |
+
},
|
2557 |
+
{
|
2558 |
+
"epoch": 92.93598233995586,
|
2559 |
+
"learning_rate": 8.711149825783971e-07,
|
2560 |
+
"loss": 1.4127,
|
2561 |
+
"step": 8420
|
2562 |
+
},
|
2563 |
+
{
|
2564 |
+
"epoch": 93.15673289183223,
|
2565 |
+
"learning_rate": 8.708013937282229e-07,
|
2566 |
+
"loss": 1.4264,
|
2567 |
+
"step": 8440
|
2568 |
+
},
|
2569 |
+
{
|
2570 |
+
"epoch": 93.37748344370861,
|
2571 |
+
"learning_rate": 8.704878048780487e-07,
|
2572 |
+
"loss": 1.4663,
|
2573 |
+
"step": 8460
|
2574 |
+
},
|
2575 |
+
{
|
2576 |
+
"epoch": 93.59823399558499,
|
2577 |
+
"learning_rate": 8.701742160278746e-07,
|
2578 |
+
"loss": 1.4151,
|
2579 |
+
"step": 8480
|
2580 |
+
},
|
2581 |
+
{
|
2582 |
+
"epoch": 93.81898454746137,
|
2583 |
+
"learning_rate": 8.698606271777002e-07,
|
2584 |
+
"loss": 1.4362,
|
2585 |
+
"step": 8500
|
2586 |
+
},
|
2587 |
+
{
|
2588 |
+
"epoch": 94.03973509933775,
|
2589 |
+
"learning_rate": 8.695470383275261e-07,
|
2590 |
+
"loss": 1.3755,
|
2591 |
+
"step": 8520
|
2592 |
+
},
|
2593 |
+
{
|
2594 |
+
"epoch": 94.26048565121413,
|
2595 |
+
"learning_rate": 8.69233449477352e-07,
|
2596 |
+
"loss": 1.3927,
|
2597 |
+
"step": 8540
|
2598 |
+
},
|
2599 |
+
{
|
2600 |
+
"epoch": 94.4812362030905,
|
2601 |
+
"learning_rate": 8.689198606271777e-07,
|
2602 |
+
"loss": 1.4137,
|
2603 |
+
"step": 8560
|
2604 |
+
},
|
2605 |
+
{
|
2606 |
+
"epoch": 94.70198675496688,
|
2607 |
+
"learning_rate": 8.686062717770035e-07,
|
2608 |
+
"loss": 1.4284,
|
2609 |
+
"step": 8580
|
2610 |
+
},
|
2611 |
+
{
|
2612 |
+
"epoch": 94.92273730684327,
|
2613 |
+
"learning_rate": 8.682926829268292e-07,
|
2614 |
+
"loss": 1.4287,
|
2615 |
+
"step": 8600
|
2616 |
+
},
|
2617 |
+
{
|
2618 |
+
"epoch": 95.14348785871965,
|
2619 |
+
"learning_rate": 8.679790940766551e-07,
|
2620 |
+
"loss": 1.4243,
|
2621 |
+
"step": 8620
|
2622 |
+
},
|
2623 |
+
{
|
2624 |
+
"epoch": 95.36423841059603,
|
2625 |
+
"learning_rate": 8.676655052264807e-07,
|
2626 |
+
"loss": 1.3588,
|
2627 |
+
"step": 8640
|
2628 |
+
},
|
2629 |
+
{
|
2630 |
+
"epoch": 95.58498896247241,
|
2631 |
+
"learning_rate": 8.673519163763066e-07,
|
2632 |
+
"loss": 1.3846,
|
2633 |
+
"step": 8660
|
2634 |
+
},
|
2635 |
+
{
|
2636 |
+
"epoch": 95.80573951434879,
|
2637 |
+
"learning_rate": 8.670383275261325e-07,
|
2638 |
+
"loss": 1.4513,
|
2639 |
+
"step": 8680
|
2640 |
+
},
|
2641 |
+
{
|
2642 |
+
"epoch": 96.02649006622516,
|
2643 |
+
"learning_rate": 8.667247386759582e-07,
|
2644 |
+
"loss": 1.4022,
|
2645 |
+
"step": 8700
|
2646 |
+
},
|
2647 |
+
{
|
2648 |
+
"epoch": 96.24724061810154,
|
2649 |
+
"learning_rate": 8.664111498257838e-07,
|
2650 |
+
"loss": 1.3969,
|
2651 |
+
"step": 8720
|
2652 |
+
},
|
2653 |
+
{
|
2654 |
+
"epoch": 96.46799116997792,
|
2655 |
+
"learning_rate": 8.660975609756097e-07,
|
2656 |
+
"loss": 1.38,
|
2657 |
+
"step": 8740
|
2658 |
+
},
|
2659 |
+
{
|
2660 |
+
"epoch": 96.6887417218543,
|
2661 |
+
"learning_rate": 8.657839721254355e-07,
|
2662 |
+
"loss": 1.4701,
|
2663 |
+
"step": 8760
|
2664 |
+
},
|
2665 |
+
{
|
2666 |
+
"epoch": 96.90949227373068,
|
2667 |
+
"learning_rate": 8.654703832752613e-07,
|
2668 |
+
"loss": 1.3637,
|
2669 |
+
"step": 8780
|
2670 |
+
},
|
2671 |
+
{
|
2672 |
+
"epoch": 97.13024282560707,
|
2673 |
+
"learning_rate": 8.65156794425087e-07,
|
2674 |
+
"loss": 1.3748,
|
2675 |
+
"step": 8800
|
2676 |
+
},
|
2677 |
+
{
|
2678 |
+
"epoch": 97.35099337748345,
|
2679 |
+
"learning_rate": 8.648432055749129e-07,
|
2680 |
+
"loss": 1.461,
|
2681 |
+
"step": 8820
|
2682 |
+
},
|
2683 |
+
{
|
2684 |
+
"epoch": 97.57174392935983,
|
2685 |
+
"learning_rate": 8.645296167247387e-07,
|
2686 |
+
"loss": 1.4615,
|
2687 |
+
"step": 8840
|
2688 |
+
},
|
2689 |
+
{
|
2690 |
+
"epoch": 97.7924944812362,
|
2691 |
+
"learning_rate": 8.642160278745644e-07,
|
2692 |
+
"loss": 1.3775,
|
2693 |
+
"step": 8860
|
2694 |
+
},
|
2695 |
+
{
|
2696 |
+
"epoch": 98.01324503311258,
|
2697 |
+
"learning_rate": 8.639024390243902e-07,
|
2698 |
+
"loss": 1.4125,
|
2699 |
+
"step": 8880
|
2700 |
+
},
|
2701 |
+
{
|
2702 |
+
"epoch": 98.23399558498896,
|
2703 |
+
"learning_rate": 8.63588850174216e-07,
|
2704 |
+
"loss": 1.3846,
|
2705 |
+
"step": 8900
|
2706 |
+
},
|
2707 |
+
{
|
2708 |
+
"epoch": 98.45474613686534,
|
2709 |
+
"learning_rate": 8.632752613240419e-07,
|
2710 |
+
"loss": 1.3948,
|
2711 |
+
"step": 8920
|
2712 |
+
},
|
2713 |
+
{
|
2714 |
+
"epoch": 98.67549668874172,
|
2715 |
+
"learning_rate": 8.629616724738675e-07,
|
2716 |
+
"loss": 1.348,
|
2717 |
+
"step": 8940
|
2718 |
+
},
|
2719 |
+
{
|
2720 |
+
"epoch": 98.8962472406181,
|
2721 |
+
"learning_rate": 8.626480836236934e-07,
|
2722 |
+
"loss": 1.4504,
|
2723 |
+
"step": 8960
|
2724 |
+
},
|
2725 |
+
{
|
2726 |
+
"epoch": 99.11699779249449,
|
2727 |
+
"learning_rate": 8.623344947735191e-07,
|
2728 |
+
"loss": 1.3899,
|
2729 |
+
"step": 8980
|
2730 |
+
},
|
2731 |
+
{
|
2732 |
+
"epoch": 99.33774834437087,
|
2733 |
+
"learning_rate": 8.62020905923345e-07,
|
2734 |
+
"loss": 1.4558,
|
2735 |
+
"step": 9000
|
2736 |
+
},
|
2737 |
+
{
|
2738 |
+
"epoch": 99.55849889624724,
|
2739 |
+
"learning_rate": 8.617073170731706e-07,
|
2740 |
+
"loss": 1.3781,
|
2741 |
+
"step": 9020
|
2742 |
+
},
|
2743 |
+
{
|
2744 |
+
"epoch": 99.77924944812362,
|
2745 |
+
"learning_rate": 8.613937282229965e-07,
|
2746 |
+
"loss": 1.3847,
|
2747 |
+
"step": 9040
|
2748 |
+
},
|
2749 |
+
{
|
2750 |
+
"epoch": 100.0,
|
2751 |
+
"learning_rate": 8.610801393728222e-07,
|
2752 |
+
"loss": 1.4043,
|
2753 |
+
"step": 9060
|
2754 |
+
},
|
2755 |
+
{
|
2756 |
+
"epoch": 100.22075055187638,
|
2757 |
+
"learning_rate": 8.60766550522648e-07,
|
2758 |
+
"loss": 1.4069,
|
2759 |
+
"step": 9080
|
2760 |
+
},
|
2761 |
+
{
|
2762 |
+
"epoch": 100.44150110375276,
|
2763 |
+
"learning_rate": 8.604529616724739e-07,
|
2764 |
+
"loss": 1.3562,
|
2765 |
+
"step": 9100
|
2766 |
+
},
|
2767 |
+
{
|
2768 |
+
"epoch": 100.66225165562913,
|
2769 |
+
"learning_rate": 8.601393728222996e-07,
|
2770 |
+
"loss": 1.3854,
|
2771 |
+
"step": 9120
|
2772 |
+
},
|
2773 |
+
{
|
2774 |
+
"epoch": 100.88300220750551,
|
2775 |
+
"learning_rate": 8.598257839721255e-07,
|
2776 |
+
"loss": 1.3842,
|
2777 |
+
"step": 9140
|
2778 |
+
},
|
2779 |
+
{
|
2780 |
+
"epoch": 101.1037527593819,
|
2781 |
+
"learning_rate": 8.595121951219512e-07,
|
2782 |
+
"loss": 1.3954,
|
2783 |
+
"step": 9160
|
2784 |
+
},
|
2785 |
+
{
|
2786 |
+
"epoch": 101.32450331125828,
|
2787 |
+
"learning_rate": 8.59198606271777e-07,
|
2788 |
+
"loss": 1.3966,
|
2789 |
+
"step": 9180
|
2790 |
+
},
|
2791 |
+
{
|
2792 |
+
"epoch": 101.54525386313466,
|
2793 |
+
"learning_rate": 8.588850174216027e-07,
|
2794 |
+
"loss": 1.4175,
|
2795 |
+
"step": 9200
|
2796 |
+
},
|
2797 |
+
{
|
2798 |
+
"epoch": 101.76600441501104,
|
2799 |
+
"learning_rate": 8.585714285714286e-07,
|
2800 |
+
"loss": 1.3944,
|
2801 |
+
"step": 9220
|
2802 |
+
},
|
2803 |
+
{
|
2804 |
+
"epoch": 101.98675496688742,
|
2805 |
+
"learning_rate": 8.582578397212543e-07,
|
2806 |
+
"loss": 1.343,
|
2807 |
+
"step": 9240
|
2808 |
+
},
|
2809 |
+
{
|
2810 |
+
"epoch": 102.2075055187638,
|
2811 |
+
"learning_rate": 8.579442508710801e-07,
|
2812 |
+
"loss": 1.3974,
|
2813 |
+
"step": 9260
|
2814 |
+
},
|
2815 |
+
{
|
2816 |
+
"epoch": 102.42825607064017,
|
2817 |
+
"learning_rate": 8.576306620209059e-07,
|
2818 |
+
"loss": 1.4235,
|
2819 |
+
"step": 9280
|
2820 |
+
},
|
2821 |
+
{
|
2822 |
+
"epoch": 102.64900662251655,
|
2823 |
+
"learning_rate": 8.573170731707317e-07,
|
2824 |
+
"loss": 1.379,
|
2825 |
+
"step": 9300
|
2826 |
+
},
|
2827 |
+
{
|
2828 |
+
"epoch": 102.86975717439293,
|
2829 |
+
"learning_rate": 8.570034843205574e-07,
|
2830 |
+
"loss": 1.3593,
|
2831 |
+
"step": 9320
|
2832 |
+
},
|
2833 |
+
{
|
2834 |
+
"epoch": 103.09050772626932,
|
2835 |
+
"learning_rate": 8.566898954703832e-07,
|
2836 |
+
"loss": 1.4164,
|
2837 |
+
"step": 9340
|
2838 |
+
},
|
2839 |
+
{
|
2840 |
+
"epoch": 103.3112582781457,
|
2841 |
+
"learning_rate": 8.563763066202091e-07,
|
2842 |
+
"loss": 1.3811,
|
2843 |
+
"step": 9360
|
2844 |
+
},
|
2845 |
+
{
|
2846 |
+
"epoch": 103.53200883002208,
|
2847 |
+
"learning_rate": 8.560627177700348e-07,
|
2848 |
+
"loss": 1.3853,
|
2849 |
+
"step": 9380
|
2850 |
+
},
|
2851 |
+
{
|
2852 |
+
"epoch": 103.75275938189846,
|
2853 |
+
"learning_rate": 8.557491289198606e-07,
|
2854 |
+
"loss": 1.3948,
|
2855 |
+
"step": 9400
|
2856 |
+
},
|
2857 |
+
{
|
2858 |
+
"epoch": 103.97350993377484,
|
2859 |
+
"learning_rate": 8.554355400696864e-07,
|
2860 |
+
"loss": 1.4206,
|
2861 |
+
"step": 9420
|
2862 |
+
},
|
2863 |
+
{
|
2864 |
+
"epoch": 104.19426048565121,
|
2865 |
+
"learning_rate": 8.551219512195122e-07,
|
2866 |
+
"loss": 1.3967,
|
2867 |
+
"step": 9440
|
2868 |
+
},
|
2869 |
+
{
|
2870 |
+
"epoch": 104.41501103752759,
|
2871 |
+
"learning_rate": 8.548083623693379e-07,
|
2872 |
+
"loss": 1.3597,
|
2873 |
+
"step": 9460
|
2874 |
+
},
|
2875 |
+
{
|
2876 |
+
"epoch": 104.63576158940397,
|
2877 |
+
"learning_rate": 8.544947735191637e-07,
|
2878 |
+
"loss": 1.4107,
|
2879 |
+
"step": 9480
|
2880 |
+
},
|
2881 |
+
{
|
2882 |
+
"epoch": 104.85651214128035,
|
2883 |
+
"learning_rate": 8.541811846689896e-07,
|
2884 |
+
"loss": 1.3626,
|
2885 |
+
"step": 9500
|
2886 |
+
},
|
2887 |
+
{
|
2888 |
+
"epoch": 105.07726269315673,
|
2889 |
+
"learning_rate": 8.538675958188153e-07,
|
2890 |
+
"loss": 1.4051,
|
2891 |
+
"step": 9520
|
2892 |
+
},
|
2893 |
+
{
|
2894 |
+
"epoch": 105.29801324503312,
|
2895 |
+
"learning_rate": 8.53554006968641e-07,
|
2896 |
+
"loss": 1.4063,
|
2897 |
+
"step": 9540
|
2898 |
+
},
|
2899 |
+
{
|
2900 |
+
"epoch": 105.5187637969095,
|
2901 |
+
"learning_rate": 8.532404181184669e-07,
|
2902 |
+
"loss": 1.4275,
|
2903 |
+
"step": 9560
|
2904 |
+
},
|
2905 |
+
{
|
2906 |
+
"epoch": 105.73951434878587,
|
2907 |
+
"learning_rate": 8.529268292682926e-07,
|
2908 |
+
"loss": 1.3462,
|
2909 |
+
"step": 9580
|
2910 |
+
},
|
2911 |
+
{
|
2912 |
+
"epoch": 105.96026490066225,
|
2913 |
+
"learning_rate": 8.526132404181184e-07,
|
2914 |
+
"loss": 1.3702,
|
2915 |
+
"step": 9600
|
2916 |
+
},
|
2917 |
+
{
|
2918 |
+
"epoch": 106.18101545253863,
|
2919 |
+
"learning_rate": 8.522996515679441e-07,
|
2920 |
+
"loss": 1.3683,
|
2921 |
+
"step": 9620
|
2922 |
+
},
|
2923 |
+
{
|
2924 |
+
"epoch": 106.40176600441501,
|
2925 |
+
"learning_rate": 8.5198606271777e-07,
|
2926 |
+
"loss": 1.4073,
|
2927 |
+
"step": 9640
|
2928 |
+
},
|
2929 |
+
{
|
2930 |
+
"epoch": 106.62251655629139,
|
2931 |
+
"learning_rate": 8.516724738675959e-07,
|
2932 |
+
"loss": 1.3627,
|
2933 |
+
"step": 9660
|
2934 |
+
},
|
2935 |
+
{
|
2936 |
+
"epoch": 106.84326710816777,
|
2937 |
+
"learning_rate": 8.513588850174215e-07,
|
2938 |
+
"loss": 1.4,
|
2939 |
+
"step": 9680
|
2940 |
+
},
|
2941 |
+
{
|
2942 |
+
"epoch": 107.06401766004414,
|
2943 |
+
"learning_rate": 8.510452961672474e-07,
|
2944 |
+
"loss": 1.3322,
|
2945 |
+
"step": 9700
|
2946 |
+
},
|
2947 |
+
{
|
2948 |
+
"epoch": 107.28476821192054,
|
2949 |
+
"learning_rate": 8.507317073170731e-07,
|
2950 |
+
"loss": 1.3776,
|
2951 |
+
"step": 9720
|
2952 |
+
},
|
2953 |
+
{
|
2954 |
+
"epoch": 107.50551876379691,
|
2955 |
+
"learning_rate": 8.50418118466899e-07,
|
2956 |
+
"loss": 1.3837,
|
2957 |
+
"step": 9740
|
2958 |
+
},
|
2959 |
+
{
|
2960 |
+
"epoch": 107.72626931567329,
|
2961 |
+
"learning_rate": 8.501045296167246e-07,
|
2962 |
+
"loss": 1.3633,
|
2963 |
+
"step": 9760
|
2964 |
+
},
|
2965 |
+
{
|
2966 |
+
"epoch": 107.94701986754967,
|
2967 |
+
"learning_rate": 8.497909407665505e-07,
|
2968 |
+
"loss": 1.4582,
|
2969 |
+
"step": 9780
|
2970 |
+
},
|
2971 |
+
{
|
2972 |
+
"epoch": 108.16777041942605,
|
2973 |
+
"learning_rate": 8.494773519163763e-07,
|
2974 |
+
"loss": 1.4204,
|
2975 |
+
"step": 9800
|
2976 |
+
},
|
2977 |
+
{
|
2978 |
+
"epoch": 108.38852097130243,
|
2979 |
+
"learning_rate": 8.49163763066202e-07,
|
2980 |
+
"loss": 1.3606,
|
2981 |
+
"step": 9820
|
2982 |
+
},
|
2983 |
+
{
|
2984 |
+
"epoch": 108.6092715231788,
|
2985 |
+
"learning_rate": 8.488501742160278e-07,
|
2986 |
+
"loss": 1.41,
|
2987 |
+
"step": 9840
|
2988 |
+
},
|
2989 |
+
{
|
2990 |
+
"epoch": 108.83002207505518,
|
2991 |
+
"learning_rate": 8.485365853658536e-07,
|
2992 |
+
"loss": 1.321,
|
2993 |
+
"step": 9860
|
2994 |
+
},
|
2995 |
+
{
|
2996 |
+
"epoch": 109.05077262693156,
|
2997 |
+
"learning_rate": 8.482229965156795e-07,
|
2998 |
+
"loss": 1.3471,
|
2999 |
+
"step": 9880
|
3000 |
+
},
|
3001 |
+
{
|
3002 |
+
"epoch": 109.27152317880795,
|
3003 |
+
"learning_rate": 8.479094076655051e-07,
|
3004 |
+
"loss": 1.3809,
|
3005 |
+
"step": 9900
|
3006 |
+
},
|
3007 |
+
{
|
3008 |
+
"epoch": 109.49227373068433,
|
3009 |
+
"learning_rate": 8.47595818815331e-07,
|
3010 |
+
"loss": 1.3795,
|
3011 |
+
"step": 9920
|
3012 |
+
},
|
3013 |
+
{
|
3014 |
+
"epoch": 109.71302428256071,
|
3015 |
+
"learning_rate": 8.472822299651568e-07,
|
3016 |
+
"loss": 1.3751,
|
3017 |
+
"step": 9940
|
3018 |
+
},
|
3019 |
+
{
|
3020 |
+
"epoch": 109.93377483443709,
|
3021 |
+
"learning_rate": 8.469686411149826e-07,
|
3022 |
+
"loss": 1.3513,
|
3023 |
+
"step": 9960
|
3024 |
+
},
|
3025 |
+
{
|
3026 |
+
"epoch": 110.15452538631347,
|
3027 |
+
"learning_rate": 8.466550522648083e-07,
|
3028 |
+
"loss": 1.3526,
|
3029 |
+
"step": 9980
|
3030 |
+
},
|
3031 |
+
{
|
3032 |
+
"epoch": 110.37527593818984,
|
3033 |
+
"learning_rate": 8.463414634146341e-07,
|
3034 |
+
"loss": 1.3819,
|
3035 |
+
"step": 10000
|
3036 |
+
},
|
3037 |
+
{
|
3038 |
+
"epoch": 110.59602649006622,
|
3039 |
+
"learning_rate": 8.460278745644599e-07,
|
3040 |
+
"loss": 1.3961,
|
3041 |
+
"step": 10020
|
3042 |
+
},
|
3043 |
+
{
|
3044 |
+
"epoch": 110.8167770419426,
|
3045 |
+
"learning_rate": 8.457142857142856e-07,
|
3046 |
+
"loss": 1.3934,
|
3047 |
+
"step": 10040
|
3048 |
+
},
|
3049 |
+
{
|
3050 |
+
"epoch": 111.03752759381898,
|
3051 |
+
"learning_rate": 8.454006968641114e-07,
|
3052 |
+
"loss": 1.3603,
|
3053 |
+
"step": 10060
|
3054 |
+
},
|
3055 |
+
{
|
3056 |
+
"epoch": 111.25827814569537,
|
3057 |
+
"learning_rate": 8.450871080139372e-07,
|
3058 |
+
"loss": 1.3933,
|
3059 |
+
"step": 10080
|
3060 |
+
},
|
3061 |
+
{
|
3062 |
+
"epoch": 111.47902869757175,
|
3063 |
+
"learning_rate": 8.44773519163763e-07,
|
3064 |
+
"loss": 1.381,
|
3065 |
+
"step": 10100
|
3066 |
+
},
|
3067 |
+
{
|
3068 |
+
"epoch": 111.69977924944813,
|
3069 |
+
"learning_rate": 8.444599303135888e-07,
|
3070 |
+
"loss": 1.4153,
|
3071 |
+
"step": 10120
|
3072 |
+
},
|
3073 |
+
{
|
3074 |
+
"epoch": 111.9205298013245,
|
3075 |
+
"learning_rate": 8.441463414634147e-07,
|
3076 |
+
"loss": 1.361,
|
3077 |
+
"step": 10140
|
3078 |
+
},
|
3079 |
+
{
|
3080 |
+
"epoch": 112.14128035320088,
|
3081 |
+
"learning_rate": 8.438327526132404e-07,
|
3082 |
+
"loss": 1.3341,
|
3083 |
+
"step": 10160
|
3084 |
+
},
|
3085 |
+
{
|
3086 |
+
"epoch": 112.36203090507726,
|
3087 |
+
"learning_rate": 8.435191637630662e-07,
|
3088 |
+
"loss": 1.3696,
|
3089 |
+
"step": 10180
|
3090 |
+
},
|
3091 |
+
{
|
3092 |
+
"epoch": 112.58278145695364,
|
3093 |
+
"learning_rate": 8.432055749128919e-07,
|
3094 |
+
"loss": 1.3899,
|
3095 |
+
"step": 10200
|
3096 |
+
},
|
3097 |
+
{
|
3098 |
+
"epoch": 112.80353200883002,
|
3099 |
+
"learning_rate": 8.428919860627178e-07,
|
3100 |
+
"loss": 1.3883,
|
3101 |
+
"step": 10220
|
3102 |
+
},
|
3103 |
+
{
|
3104 |
+
"epoch": 113.0242825607064,
|
3105 |
+
"learning_rate": 8.425783972125435e-07,
|
3106 |
+
"loss": 1.3401,
|
3107 |
+
"step": 10240
|
3108 |
+
},
|
3109 |
+
{
|
3110 |
+
"epoch": 113.24503311258277,
|
3111 |
+
"learning_rate": 8.422648083623693e-07,
|
3112 |
+
"loss": 1.3599,
|
3113 |
+
"step": 10260
|
3114 |
+
},
|
3115 |
+
{
|
3116 |
+
"epoch": 113.46578366445917,
|
3117 |
+
"learning_rate": 8.419512195121951e-07,
|
3118 |
+
"loss": 1.3688,
|
3119 |
+
"step": 10280
|
3120 |
+
},
|
3121 |
+
{
|
3122 |
+
"epoch": 113.68653421633555,
|
3123 |
+
"learning_rate": 8.416376306620209e-07,
|
3124 |
+
"loss": 1.3976,
|
3125 |
+
"step": 10300
|
3126 |
+
},
|
3127 |
+
{
|
3128 |
+
"epoch": 113.90728476821192,
|
3129 |
+
"learning_rate": 8.413240418118465e-07,
|
3130 |
+
"loss": 1.3909,
|
3131 |
+
"step": 10320
|
3132 |
+
},
|
3133 |
+
{
|
3134 |
+
"epoch": 114.1280353200883,
|
3135 |
+
"learning_rate": 8.410104529616724e-07,
|
3136 |
+
"loss": 1.4132,
|
3137 |
+
"step": 10340
|
3138 |
+
},
|
3139 |
+
{
|
3140 |
+
"epoch": 114.34878587196468,
|
3141 |
+
"learning_rate": 8.406968641114982e-07,
|
3142 |
+
"loss": 1.3741,
|
3143 |
+
"step": 10360
|
3144 |
+
},
|
3145 |
+
{
|
3146 |
+
"epoch": 114.56953642384106,
|
3147 |
+
"learning_rate": 8.40383275261324e-07,
|
3148 |
+
"loss": 1.3339,
|
3149 |
+
"step": 10380
|
3150 |
+
},
|
3151 |
+
{
|
3152 |
+
"epoch": 114.79028697571744,
|
3153 |
+
"learning_rate": 8.400696864111498e-07,
|
3154 |
+
"loss": 1.3869,
|
3155 |
+
"step": 10400
|
3156 |
+
},
|
3157 |
+
{
|
3158 |
+
"epoch": 115.01103752759381,
|
3159 |
+
"learning_rate": 8.397560975609756e-07,
|
3160 |
+
"loss": 1.3644,
|
3161 |
+
"step": 10420
|
3162 |
+
},
|
3163 |
+
{
|
3164 |
+
"epoch": 115.23178807947019,
|
3165 |
+
"learning_rate": 8.394425087108014e-07,
|
3166 |
+
"loss": 1.3584,
|
3167 |
+
"step": 10440
|
3168 |
+
},
|
3169 |
+
{
|
3170 |
+
"epoch": 115.45253863134658,
|
3171 |
+
"learning_rate": 8.391289198606271e-07,
|
3172 |
+
"loss": 1.3698,
|
3173 |
+
"step": 10460
|
3174 |
+
},
|
3175 |
+
{
|
3176 |
+
"epoch": 115.67328918322296,
|
3177 |
+
"learning_rate": 8.388153310104529e-07,
|
3178 |
+
"loss": 1.39,
|
3179 |
+
"step": 10480
|
3180 |
+
},
|
3181 |
+
{
|
3182 |
+
"epoch": 115.89403973509934,
|
3183 |
+
"learning_rate": 8.385017421602787e-07,
|
3184 |
+
"loss": 1.3651,
|
3185 |
+
"step": 10500
|
3186 |
+
},
|
3187 |
+
{
|
3188 |
+
"epoch": 116.11479028697572,
|
3189 |
+
"learning_rate": 8.381881533101045e-07,
|
3190 |
+
"loss": 1.339,
|
3191 |
+
"step": 10520
|
3192 |
+
},
|
3193 |
+
{
|
3194 |
+
"epoch": 116.3355408388521,
|
3195 |
+
"learning_rate": 8.378745644599303e-07,
|
3196 |
+
"loss": 1.3469,
|
3197 |
+
"step": 10540
|
3198 |
+
},
|
3199 |
+
{
|
3200 |
+
"epoch": 116.55629139072848,
|
3201 |
+
"learning_rate": 8.375609756097561e-07,
|
3202 |
+
"loss": 1.4008,
|
3203 |
+
"step": 10560
|
3204 |
+
},
|
3205 |
+
{
|
3206 |
+
"epoch": 116.77704194260485,
|
3207 |
+
"learning_rate": 8.372473867595818e-07,
|
3208 |
+
"loss": 1.3757,
|
3209 |
+
"step": 10580
|
3210 |
+
},
|
3211 |
+
{
|
3212 |
+
"epoch": 116.99779249448123,
|
3213 |
+
"learning_rate": 8.369337979094076e-07,
|
3214 |
+
"loss": 1.3792,
|
3215 |
+
"step": 10600
|
3216 |
+
},
|
3217 |
+
{
|
3218 |
+
"epoch": 117.21854304635761,
|
3219 |
+
"learning_rate": 8.366202090592334e-07,
|
3220 |
+
"loss": 1.3422,
|
3221 |
+
"step": 10620
|
3222 |
+
},
|
3223 |
+
{
|
3224 |
+
"epoch": 117.439293598234,
|
3225 |
+
"learning_rate": 8.363066202090592e-07,
|
3226 |
+
"loss": 1.3876,
|
3227 |
+
"step": 10640
|
3228 |
+
},
|
3229 |
+
{
|
3230 |
+
"epoch": 117.66004415011038,
|
3231 |
+
"learning_rate": 8.359930313588849e-07,
|
3232 |
+
"loss": 1.3519,
|
3233 |
+
"step": 10660
|
3234 |
+
},
|
3235 |
+
{
|
3236 |
+
"epoch": 117.88079470198676,
|
3237 |
+
"learning_rate": 8.356794425087108e-07,
|
3238 |
+
"loss": 1.3268,
|
3239 |
+
"step": 10680
|
3240 |
+
},
|
3241 |
+
{
|
3242 |
+
"epoch": 118.10154525386314,
|
3243 |
+
"learning_rate": 8.353658536585366e-07,
|
3244 |
+
"loss": 1.4245,
|
3245 |
+
"step": 10700
|
3246 |
+
},
|
3247 |
+
{
|
3248 |
+
"epoch": 118.32229580573951,
|
3249 |
+
"learning_rate": 8.350522648083623e-07,
|
3250 |
+
"loss": 1.3755,
|
3251 |
+
"step": 10720
|
3252 |
+
},
|
3253 |
+
{
|
3254 |
+
"epoch": 118.54304635761589,
|
3255 |
+
"learning_rate": 8.347386759581881e-07,
|
3256 |
+
"loss": 1.3318,
|
3257 |
+
"step": 10740
|
3258 |
+
},
|
3259 |
+
{
|
3260 |
+
"epoch": 118.76379690949227,
|
3261 |
+
"learning_rate": 8.344250871080139e-07,
|
3262 |
+
"loss": 1.3395,
|
3263 |
+
"step": 10760
|
3264 |
+
},
|
3265 |
+
{
|
3266 |
+
"epoch": 118.98454746136865,
|
3267 |
+
"learning_rate": 8.341114982578397e-07,
|
3268 |
+
"loss": 1.407,
|
3269 |
+
"step": 10780
|
3270 |
+
},
|
3271 |
+
{
|
3272 |
+
"epoch": 119.20529801324503,
|
3273 |
+
"learning_rate": 8.337979094076654e-07,
|
3274 |
+
"loss": 1.3621,
|
3275 |
+
"step": 10800
|
3276 |
+
},
|
3277 |
+
{
|
3278 |
+
"epoch": 119.20529801324503,
|
3279 |
+
"eval_bleu": 49.121,
|
3280 |
+
"eval_gen_len": 8.8,
|
3281 |
+
"eval_loss": 2.0037317276000977,
|
3282 |
+
"eval_runtime": 3.2416,
|
3283 |
+
"eval_samples_per_second": 9.255,
|
3284 |
+
"eval_steps_per_second": 1.851,
|
3285 |
+
"step": 10800
|
3286 |
}
|
3287 |
],
|
3288 |
"logging_steps": 20,
|
|
|
3302 |
"attributes": {}
|
3303 |
}
|
3304 |
},
|
3305 |
+
"total_flos": 3.194864182891315e+16,
|
3306 |
"train_batch_size": 5,
|
3307 |
"trial_name": null,
|
3308 |
"trial_params": null
|