gama-it

app.py

@@ -38,10 +38,7 @@ temp, top_p, top_k = 0.1, 0.95, 500
 # change it to your model path
 
 ### Stage 4 ckpt
-eval_mdl_path = './stage4_ckpt/pytorch_model.bin'
-
-### Stage 5 ckpt
-# eval_mdl_path = '/fs/gamma-projects/audio/gama/new_data/stage5_all_mix_all/checkpoint-900/pytorch_model.bin'
+eval_mdl_path = './stage5_ckpt/pytorch_model.bin'
 
 state_dict = torch.load(eval_mdl_path, map_location='cpu')
 msg = model.load_state_dict(state_dict, strict=False)
@@ -146,12 +143,12 @@ text = "[Github]"
 paper_link = "https://sreyan88.github.io/gamaaudio/"
 paper_text = "[Paper]"
 demo = gr.Interface(fn=predict,
-                    inputs=[gr.Audio(type="filepath"), gr.Textbox(value='Describe the audio.', label='Edit the textbox to ask your own questions!')],
+                    inputs=[gr.Audio(type="filepath"), gr.Textbox(value='Describe the audio in detail', label='Edit the textbox to ask your own questions!')],
                     outputs=[gr.Textbox(label="Audio Meta Information"), gr.Textbox(label="GAMA Output")],
                     cache_examples=True,
-                    title="Quick Demo of GAMA",
-                    description="GAMA is a novel Large Large Audio-Language Model that is capable of understanding audio inputs and answer any open-ended question about it." + f"<a href='{paper_link}'>{paper_text}</a> " + f"<a href='{link}'>{text}</a> <br>" +
-                    "GAMA is authored by members of the GAMMA Lab at the University of Maryland, College Park and Adobe, USA. <br>" +
-                    "**GAMA is not an ASR model and has limited ability to recognize the speech content. It primarily focuses on perception and understanding of non-speech sounds.**<br>" +
+                    title="Quick Demo of GAMA-IT",
+                    description="GAMA-IT is an instruction-tuned version of GAMA, a novel Large Audio-Language Model that is capable of complex reasoning. The model can understand any given audio (non-speech) and answer any open-ended question about it, including ones that require complex reasoning." + f"<a href='{paper_link}'>{paper_text}</a> " + f"<a href='{link}'>{text}</a> <br>" +
+                    "GAMA-IT is authored by members of the GAMMA Lab at the University of Maryland, College Park and Adobe, USA. <br>" +
+                    "**GAMA-IT is not an ASR model and has limited ability to recognize the speech content. It primarily focuses on perception and understanding of non-speech sounds.**<br>" +
                     "Input an audio and ask quesions! Audio will be converted to 16kHz and padded or trim to 10 seconds.")
 demo.launch(debug=True, share=True)

stage4_ckpt/rng_state_4.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:3981a483a45df0d096dc359bf36d393f20c8949ba758a031ead466753711bf5a
-size 14583

stage4_ckpt/rng_state_5.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:fbf35f8fe5f1027dc4246f14ffcccc2ba98e5871ca631a7c1c528653a3fa6ee9
-size 14583

stage4_ckpt/rng_state_6.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:0e6846f57e23b560d2a5997bf8decfc53fda9e9bc3123735157e8de1bbf3f4c7
-size 14583

stage4_ckpt/rng_state_7.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:55a4f8a650749d1b1857baf17da7e81d84caa938058eda19134d20d615582588
-size 14583

{stage4_ckpt → stage5_ckpt}/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b3e4bc35ff435c5edd17a177982a512bd1a84e1cc1e2e2acec0bd582814fc00
+oid sha256:759668b05a0262e58b6a1bd4c84e6e8ababf03f2078b0ea6cb89bf9d4e659b12
 size 1946313914

{stage4_ckpt → stage5_ckpt}/pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5a0fd007ede3e0f66203fe4525a1c7654009efd0b16c1bd13448356a1c343f1
+oid sha256:4eec0115fd316c8b5d2201d07973020d5f2f0b51bc412af88aba48d944b54246
 size 1029854941

{stage4_ckpt → stage5_ckpt}/rng_state_0.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c3431b4d910a553d68af4af283ded45ef930962b13c39bc51922ecfac8723774
+oid sha256:ea35b7d59451c22efab6bdc48407243c00af67b9a27ee027f2dc683a40f6eb80
 size 14583

{stage4_ckpt → stage5_ckpt}/rng_state_1.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c557b45aef05f633a33537da42537cb3fe314438f647dad2736fb44fd264010e
+oid sha256:49cd382683c500be01985d3e70cda9e23083436b29e6cb7aeb4e069f1feeebe1
 size 14583

{stage4_ckpt → stage5_ckpt}/rng_state_2.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:144d1b40e29e2e1f6a63f4216aad119210eaca48b06d11935059395e72321078
+oid sha256:aac490b7026c56722972fc3c077397c72daefc97030f6cf98533820ea6074fa3
 size 14583

{stage4_ckpt → stage5_ckpt}/rng_state_3.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2e1eaab914231dfc8eed2460f9fda2bca54c999e19aed14d0ca219255e7a1a22
+oid sha256:b7386e97853e08c8e84c17f74b58f60c7aeba028318ec26a13a4f72dbe82a517
 size 14583

{stage4_ckpt → stage5_ckpt}/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:32e642db26c4cf5eb2d0447e51bd109f1bbf24ebf3c7f7473e9c30c5998702fd
+oid sha256:90e0da0bac88a4b918fa64b60525b0e983c3a691f5480f380d88f3e867c46ef0
 size 627

stage5_ckpt/trainer_state.json

@@ -0,0 +1,1516 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9988014382740711,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.8396,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.7082,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-05,
+      "loss": 1.5598,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4e-05,
+      "loss": 1.4378,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-05,
+      "loss": 1.328,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-05,
+      "loss": 1.3232,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7e-05,
+      "loss": 1.2997,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8e-05,
+      "loss": 1.2623,
+      "step": 80
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9e-05,
+      "loss": 1.2463,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001,
+      "loss": 1.2288,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958385351643778e-05,
+      "loss": 1.2473,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.916770703287557e-05,
+      "loss": 1.2466,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.875156054931336e-05,
+      "loss": 1.233,
+      "step": 130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.833541406575115e-05,
+      "loss": 1.2122,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.791926758218893e-05,
+      "loss": 1.2104,
+      "step": 150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.750312109862672e-05,
+      "loss": 1.2335,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.708697461506451e-05,
+      "loss": 1.2213,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.667082813150229e-05,
+      "loss": 1.2035,
+      "step": 180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.625468164794008e-05,
+      "loss": 1.2006,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.583853516437786e-05,
+      "loss": 1.1888,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.542238868081565e-05,
+      "loss": 1.2196,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.500624219725344e-05,
+      "loss": 1.1914,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.459009571369123e-05,
+      "loss": 1.1898,
+      "step": 230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.417394923012901e-05,
+      "loss": 1.1862,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.375780274656679e-05,
+      "loss": 1.1689,
+      "step": 250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.334165626300458e-05,
+      "loss": 1.2057,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.292550977944237e-05,
+      "loss": 1.1927,
+      "step": 270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.250936329588016e-05,
+      "loss": 1.1872,
+      "step": 280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.209321681231794e-05,
+      "loss": 1.1764,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.167707032875571e-05,
+      "loss": 1.1716,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.126092384519352e-05,
+      "loss": 1.1869,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.08447773616313e-05,
+      "loss": 1.1771,
+      "step": 320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.042863087806909e-05,
+      "loss": 1.1809,
+      "step": 330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.001248439450686e-05,
+      "loss": 1.1673,
+      "step": 340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.959633791094466e-05,
+      "loss": 1.1695,
+      "step": 350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 8.918019142738245e-05,
+      "loss": 1.1808,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.876404494382022e-05,
+      "loss": 1.1846,
+      "step": 370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.834789846025801e-05,
+      "loss": 1.1687,
+      "step": 380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.793175197669579e-05,
+      "loss": 1.1669,
+      "step": 390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.75156054931336e-05,
+      "loss": 1.1603,
+      "step": 400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 8.709945900957137e-05,
+      "loss": 1.1764,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.668331252600916e-05,
+      "loss": 1.1699,
+      "step": 420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 8.626716604244694e-05,
+      "loss": 1.1811,
+      "step": 430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.585101955888473e-05,
+      "loss": 1.1627,
+      "step": 440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.543487307532252e-05,
+      "loss": 1.1572,
+      "step": 450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 8.50187265917603e-05,
+      "loss": 1.184,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.460258010819809e-05,
+      "loss": 1.1518,
+      "step": 470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 8.418643362463587e-05,
+      "loss": 1.1847,
+      "step": 480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.377028714107366e-05,
+      "loss": 1.1757,
+      "step": 490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.335414065751145e-05,
+      "loss": 1.1387,
+      "step": 500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.293799417394923e-05,
+      "loss": 1.1592,
+      "step": 510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.252184769038702e-05,
+      "loss": 1.1635,
+      "step": 520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.21057012068248e-05,
+      "loss": 1.1531,
+      "step": 530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.16895547232626e-05,
+      "loss": 1.1423,
+      "step": 540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.127340823970038e-05,
+      "loss": 1.1306,
+      "step": 550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.085726175613817e-05,
+      "loss": 1.1808,
+      "step": 560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.044111527257595e-05,
+      "loss": 1.1521,
+      "step": 570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.002496878901374e-05,
+      "loss": 1.1473,
+      "step": 580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.960882230545153e-05,
+      "loss": 1.1507,
+      "step": 590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.91926758218893e-05,
+      "loss": 1.1421,
+      "step": 600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.87765293383271e-05,
+      "loss": 1.1786,
+      "step": 610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.836038285476487e-05,
+      "loss": 1.1665,
+      "step": 620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.794423637120268e-05,
+      "loss": 1.1514,
+      "step": 630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.752808988764046e-05,
+      "loss": 1.1383,
+      "step": 640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.711194340407823e-05,
+      "loss": 1.1415,
+      "step": 650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.669579692051602e-05,
+      "loss": 1.1586,
+      "step": 660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.627965043695382e-05,
+      "loss": 1.1575,
+      "step": 670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.58635039533916e-05,
+      "loss": 1.1452,
+      "step": 680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.544735746982938e-05,
+      "loss": 1.146,
+      "step": 690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.503121098626716e-05,
+      "loss": 1.1352,
+      "step": 700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.461506450270495e-05,
+      "loss": 1.1504,
+      "step": 710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.419891801914274e-05,
+      "loss": 1.143,
+      "step": 720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.378277153558053e-05,
+      "loss": 1.155,
+      "step": 730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.336662505201831e-05,
+      "loss": 1.1334,
+      "step": 740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.29504785684561e-05,
+      "loss": 1.1228,
+      "step": 750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.253433208489388e-05,
+      "loss": 1.1548,
+      "step": 760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.211818560133167e-05,
+      "loss": 1.133,
+      "step": 770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.170203911776946e-05,
+      "loss": 1.1219,
+      "step": 780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.128589263420724e-05,
+      "loss": 1.1354,
+      "step": 790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.086974615064503e-05,
+      "loss": 1.1309,
+      "step": 800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.045359966708282e-05,
+      "loss": 1.1383,
+      "step": 810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.003745318352061e-05,
+      "loss": 1.1782,
+      "step": 820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 6.962130669995839e-05,
+      "loss": 1.1326,
+      "step": 830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.920516021639617e-05,
+      "loss": 1.1339,
+      "step": 840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.878901373283396e-05,
+      "loss": 1.1325,
+      "step": 850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 6.837286724927175e-05,
+      "loss": 1.1504,
+      "step": 860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.795672076570954e-05,
+      "loss": 1.1462,
+      "step": 870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.754057428214732e-05,
+      "loss": 1.135,
+      "step": 880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.71244277985851e-05,
+      "loss": 1.1325,
+      "step": 890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.67082813150229e-05,
+      "loss": 1.1192,
+      "step": 900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.629213483146067e-05,
+      "loss": 1.1504,
+      "step": 910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.587598834789847e-05,
+      "loss": 1.1378,
+      "step": 920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.545984186433624e-05,
+      "loss": 1.1417,
+      "step": 930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.504369538077403e-05,
+      "loss": 1.1384,
+      "step": 940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.462754889721182e-05,
+      "loss": 1.1162,
+      "step": 950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.421140241364962e-05,
+      "loss": 1.1268,
+      "step": 960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.379525593008739e-05,
+      "loss": 1.1359,
+      "step": 970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.337910944652517e-05,
+      "loss": 1.117,
+      "step": 980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.296296296296296e-05,
+      "loss": 1.1269,
+      "step": 990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.254681647940075e-05,
+      "loss": 1.1279,
+      "step": 1000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.213066999583854e-05,
+      "loss": 1.1508,
+      "step": 1010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.171452351227632e-05,
+      "loss": 1.1326,
+      "step": 1020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.129837702871411e-05,
+      "loss": 1.1325,
+      "step": 1030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.0882230545151896e-05,
+      "loss": 1.1194,
+      "step": 1040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.046608406158968e-05,
+      "loss": 1.113,
+      "step": 1050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.004993757802747e-05,
+      "loss": 1.1378,
+      "step": 1060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.9633791094465255e-05,
+      "loss": 1.1278,
+      "step": 1070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.9217644610903046e-05,
+      "loss": 1.1281,
+      "step": 1080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.880149812734082e-05,
+      "loss": 1.129,
+      "step": 1090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.838535164377861e-05,
+      "loss": 1.1265,
+      "step": 1100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.79692051602164e-05,
+      "loss": 1.1327,
+      "step": 1110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.755305867665418e-05,
+      "loss": 1.1377,
+      "step": 1120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.713691219309197e-05,
+      "loss": 1.1173,
+      "step": 1130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.672076570952976e-05,
+      "loss": 1.1231,
+      "step": 1140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.630461922596755e-05,
+      "loss": 1.1168,
+      "step": 1150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.588847274240533e-05,
+      "loss": 1.1537,
+      "step": 1160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.547232625884311e-05,
+      "loss": 1.132,
+      "step": 1170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.50561797752809e-05,
+      "loss": 1.1209,
+      "step": 1180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4640033291718685e-05,
+      "loss": 1.1139,
+      "step": 1190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4223886808156476e-05,
+      "loss": 1.1018,
+      "step": 1200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.380774032459426e-05,
+      "loss": 1.1392,
+      "step": 1210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.339159384103205e-05,
+      "loss": 1.1258,
+      "step": 1220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2975447357469835e-05,
+      "loss": 1.116,
+      "step": 1230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.255930087390761e-05,
+      "loss": 1.1326,
+      "step": 1240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.21431543903454e-05,
+      "loss": 1.1042,
+      "step": 1250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.172700790678319e-05,
+      "loss": 1.1277,
+      "step": 1260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.131086142322098e-05,
+      "loss": 1.1345,
+      "step": 1270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 5.089471493965876e-05,
+      "loss": 1.1118,
+      "step": 1280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.047856845609655e-05,
+      "loss": 1.1007,
+      "step": 1290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.006242197253434e-05,
+      "loss": 1.1147,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.964627548897212e-05,
+      "loss": 1.1251,
+      "step": 1310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.9230129005409905e-05,
+      "loss": 1.1219,
+      "step": 1320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.8813982521847696e-05,
+      "loss": 1.1191,
+      "step": 1330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.8397836038285474e-05,
+      "loss": 1.1211,
+      "step": 1340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.7981689554723265e-05,
+      "loss": 1.1006,
+      "step": 1350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.756554307116105e-05,
+      "loss": 1.1183,
+      "step": 1360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.714939658759884e-05,
+      "loss": 1.1147,
+      "step": 1370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.6733250104036624e-05,
+      "loss": 1.1212,
+      "step": 1380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.6317103620474415e-05,
+      "loss": 1.1117,
+      "step": 1390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.590095713691219e-05,
+      "loss": 1.1067,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.548481065334998e-05,
+      "loss": 1.126,
+      "step": 1410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.506866416978777e-05,
+      "loss": 1.1279,
+      "step": 1420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.465251768622555e-05,
+      "loss": 1.1147,
+      "step": 1430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.423637120266334e-05,
+      "loss": 1.1122,
+      "step": 1440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.3820224719101126e-05,
+      "loss": 1.0992,
+      "step": 1450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.340407823553892e-05,
+      "loss": 1.1174,
+      "step": 1460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2987931751976694e-05,
+      "loss": 1.1236,
+      "step": 1470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.2571785268414485e-05,
+      "loss": 1.1053,
+      "step": 1480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.215563878485227e-05,
+      "loss": 1.1091,
+      "step": 1490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1739492301290054e-05,
+      "loss": 1.1013,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.1323345817727844e-05,
+      "loss": 1.1316,
+      "step": 1510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.090719933416563e-05,
+      "loss": 1.1295,
+      "step": 1520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.049105285060342e-05,
+      "loss": 1.1178,
+      "step": 1530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.00749063670412e-05,
+      "loss": 1.106,
+      "step": 1540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.965875988347899e-05,
+      "loss": 1.0978,
+      "step": 1550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.924261339991677e-05,
+      "loss": 1.1194,
+      "step": 1560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.8826466916354556e-05,
+      "loss": 1.1136,
+      "step": 1570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.841032043279235e-05,
+      "loss": 1.111,
+      "step": 1580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.799417394923013e-05,
+      "loss": 1.1144,
+      "step": 1590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.7578027465667915e-05,
+      "loss": 1.0934,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.71618809821057e-05,
+      "loss": 1.1276,
+      "step": 1610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.674573449854349e-05,
+      "loss": 1.1121,
+      "step": 1620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.6329588014981274e-05,
+      "loss": 1.1103,
+      "step": 1630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5913441531419065e-05,
+      "loss": 1.1029,
+      "step": 1640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.549729504785685e-05,
+      "loss": 1.089,
+      "step": 1650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.5081148564294633e-05,
+      "loss": 1.1432,
+      "step": 1660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.466500208073242e-05,
+      "loss": 1.1095,
+      "step": 1670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.42488555971702e-05,
+      "loss": 1.1023,
+      "step": 1680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.383270911360799e-05,
+      "loss": 1.089,
+      "step": 1690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.341656263004578e-05,
+      "loss": 1.1064,
+      "step": 1700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.300041614648357e-05,
+      "loss": 1.1292,
+      "step": 1710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.258426966292135e-05,
+      "loss": 1.1204,
+      "step": 1720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.2168123179359136e-05,
+      "loss": 1.1058,
+      "step": 1730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.175197669579692e-05,
+      "loss": 1.1047,
+      "step": 1740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.1335830212234704e-05,
+      "loss": 1.0791,
+      "step": 1750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.0919683728672495e-05,
+      "loss": 1.1173,
+      "step": 1760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.050353724511028e-05,
+      "loss": 1.1118,
+      "step": 1770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.0087390761548067e-05,
+      "loss": 1.093,
+      "step": 1780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9671244277985854e-05,
+      "loss": 1.0979,
+      "step": 1790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.9255097794423635e-05,
+      "loss": 1.0924,
+      "step": 1800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.8838951310861422e-05,
+      "loss": 1.1274,
+      "step": 1810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.842280482729921e-05,
+      "loss": 1.1169,
+      "step": 1820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.8006658343736998e-05,
+      "loss": 1.1099,
+      "step": 1830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7590511860174785e-05,
+      "loss": 1.1,
+      "step": 1840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.717436537661257e-05,
+      "loss": 1.0847,
+      "step": 1850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.6758218893050357e-05,
+      "loss": 1.121,
+      "step": 1860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.6342072409488137e-05,
+      "loss": 1.1192,
+      "step": 1870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5925925925925925e-05,
+      "loss": 1.0986,
+      "step": 1880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5509779442363712e-05,
+      "loss": 1.1058,
+      "step": 1890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.50936329588015e-05,
+      "loss": 1.101,
+      "step": 1900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4677486475239287e-05,
+      "loss": 1.12,
+      "step": 1910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.426133999167707e-05,
+      "loss": 1.1078,
+      "step": 1920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.384519350811486e-05,
+      "loss": 1.1028,
+      "step": 1930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3429047024552643e-05,
+      "loss": 1.0957,
+      "step": 1940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3012900540990427e-05,
+      "loss": 1.0841,
+      "step": 1950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.2596754057428215e-05,
+      "loss": 1.1262,
+      "step": 1960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2180607573866002e-05,
+      "loss": 1.1049,
+      "step": 1970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.176446109030379e-05,
+      "loss": 1.0989,
+      "step": 1980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1348314606741574e-05,
+      "loss": 1.0941,
+      "step": 1990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093216812317936e-05,
+      "loss": 1.0772,
+      "step": 2000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0516021639617146e-05,
+      "loss": 1.1112,
+      "step": 2010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.009987515605493e-05,
+      "loss": 1.1162,
+      "step": 2020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9683728672492717e-05,
+      "loss": 1.1007,
+      "step": 2030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.9267582188930505e-05,
+      "loss": 1.089,
+      "step": 2040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.885143570536829e-05,
+      "loss": 1.0796,
+      "step": 2050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8435289221806076e-05,
+      "loss": 1.113,
+      "step": 2060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8019142738243864e-05,
+      "loss": 1.1066,
+      "step": 2070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.760299625468165e-05,
+      "loss": 1.099,
+      "step": 2080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.7186849771119436e-05,
+      "loss": 1.0781,
+      "step": 2090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.677070328755722e-05,
+      "loss": 1.0728,
+      "step": 2100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.6354556803995007e-05,
+      "loss": 1.1091,
+      "step": 2110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.593841032043279e-05,
+      "loss": 1.0924,
+      "step": 2120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.552226383687058e-05,
+      "loss": 1.1002,
+      "step": 2130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5106117353308366e-05,
+      "loss": 1.0892,
+      "step": 2140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.468997086974615e-05,
+      "loss": 1.0894,
+      "step": 2150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4273824386183936e-05,
+      "loss": 1.1073,
+      "step": 2160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3857677902621724e-05,
+      "loss": 1.098,
+      "step": 2170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3441531419059511e-05,
+      "loss": 1.0951,
+      "step": 2180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3025384935497296e-05,
+      "loss": 1.0867,
+      "step": 2190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2609238451935081e-05,
+      "loss": 1.0894,
+      "step": 2200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.2193091968372867e-05,
+      "loss": 1.113,
+      "step": 2210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1776945484810655e-05,
+      "loss": 1.1026,
+      "step": 2220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.136079900124844e-05,
+      "loss": 1.117,
+      "step": 2230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.0944652517686226e-05,
+      "loss": 1.0984,
+      "step": 2240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0528506034124012e-05,
+      "loss": 1.1025,
+      "step": 2250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.0112359550561798e-05,
+      "loss": 1.1252,
+      "step": 2260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.696213066999585e-06,
+      "loss": 1.0948,
+      "step": 2270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.28006658343737e-06,
+      "loss": 1.0862,
+      "step": 2280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 8.863920099875155e-06,
+      "loss": 1.0821,
+      "step": 2290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.447773616312943e-06,
+      "loss": 1.0889,
+      "step": 2300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.031627132750729e-06,
+      "loss": 1.1228,
+      "step": 2310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.615480649188515e-06,
+      "loss": 1.1051,
+      "step": 2320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.1993341656263e-06,
+      "loss": 1.0838,
+      "step": 2330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.783187682064086e-06,
+      "loss": 1.0922,
+      "step": 2340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.367041198501873e-06,
+      "loss": 1.0728,
+      "step": 2350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.950894714939659e-06,
+      "loss": 1.1049,
+      "step": 2360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.534748231377445e-06,
+      "loss": 1.0959,
+      "step": 2370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.118601747815231e-06,
+      "loss": 1.0903,
+      "step": 2380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.702455264253018e-06,
+      "loss": 1.0818,
+      "step": 2390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.286308780690804e-06,
+      "loss": 1.0704,
+      "step": 2400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.8701622971285895e-06,
+      "loss": 1.1103,
+      "step": 2410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.4540158135663753e-06,
+      "loss": 1.1007,
+      "step": 2420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.0378693300041616e-06,
+      "loss": 1.0815,
+      "step": 2430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.621722846441948e-06,
+      "loss": 1.0878,
+      "step": 2440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2055763628797336e-06,
+      "loss": 1.0896,
+      "step": 2450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7894298793175199e-06,
+      "loss": 1.1139,
+      "step": 2460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.373283395755306e-06,
+      "loss": 1.0975,
+      "step": 2470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.57136912193092e-07,
+      "loss": 1.0901,
+      "step": 2480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.409904286308781e-07,
+      "loss": 1.0836,
+      "step": 2490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2484394506866418e-07,
+      "loss": 1.0908,
+      "step": 2500
+    }
+  ],
+  "max_steps": 2503,
+  "num_train_epochs": 1,
+  "total_flos": 6.363378008362844e+17,
+  "trial_name": null,
+  "trial_params": null
+}

{stage4_ckpt → stage5_ckpt}/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a5d65d0e52e4f2e8295ff0ea14611cea8e57062d38bf8be891fe2a7bfeae9832
+oid sha256:642fdcacc3b9f137a4d752ea81ec333de641ce575b7e61ece7c885d12a21beea
 size 3771