llama-3.2-3B-lora-r2

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the gommt-oneshot-train dataset. It achieves the following results on the evaluation set:

• Loss: 0.0120

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.0578	0.0372	5	0.0572
0.0546	0.0745	10	0.0571
0.0544	0.1117	15	0.0569
0.0546	0.1490	20	0.0561
0.0584	0.1862	25	0.0540
0.0461	0.2235	30	0.0504
0.0472	0.2607	35	0.0462
0.0422	0.2980	40	0.0430
0.0354	0.3352	45	0.0408
0.038	0.3724	50	0.0386
0.034	0.4097	55	0.0363
0.0325	0.4469	60	0.0345
0.0299	0.4842	65	0.0331
0.0286	0.5214	70	0.0313
0.0299	0.5587	75	0.0292
0.028	0.5959	80	0.0276
0.025	0.6331	85	0.0261
0.024	0.6704	90	0.0248
0.0231	0.7076	95	0.0241
0.0222	0.7449	100	0.0234
0.021	0.7821	105	0.0229
0.0228	0.8194	110	0.0222
0.0184	0.8566	115	0.0216
0.0227	0.8939	120	0.0211
0.0188	0.9311	125	0.0209
0.0238	0.9683	130	0.0205
0.0328	1.0056	135	0.0201
0.0208	1.0428	140	0.0198
0.0199	1.0801	145	0.0194
0.0201	1.1173	150	0.0191
0.0163	1.1546	155	0.0189
0.0164	1.1918	160	0.0186
0.0174	1.2291	165	0.0185
0.0175	1.2663	170	0.0182
0.0175	1.3035	175	0.0180
0.0184	1.3408	180	0.0175
0.0185	1.3780	185	0.0174
0.0155	1.4153	190	0.0173
0.016	1.4525	195	0.0170
0.0191	1.4898	200	0.0167
0.0186	1.5270	205	0.0165
0.0171	1.5642	210	0.0164
0.0192	1.6015	215	0.0165
0.0154	1.6387	220	0.0159
0.0179	1.6760	225	0.0160
0.0153	1.7132	230	0.0157
0.0162	1.7505	235	0.0155
0.0166	1.7877	240	0.0154
0.0147	1.8250	245	0.0153
0.016	1.8622	250	0.0153
0.0153	1.8994	255	0.0150
0.0157	1.9367	260	0.0149
0.0165	1.9739	265	0.0150
0.0153	2.0112	270	0.0148
0.015	2.0484	275	0.0149
0.0159	2.0857	280	0.0148
0.0166	2.1229	285	0.0146
0.0153	2.1601	290	0.0146
0.013	2.1974	295	0.0143
0.0139	2.2346	300	0.0143
0.016	2.2719	305	0.0145
0.0142	2.3091	310	0.0144
0.0138	2.3464	315	0.0143
0.0151	2.3836	320	0.0143
0.0152	2.4209	325	0.0140
0.0141	2.4581	330	0.0142
0.0137	2.4953	335	0.0139
0.0132	2.5326	340	0.0138
0.0132	2.5698	345	0.0136
0.0162	2.6071	350	0.0135
0.0133	2.6443	355	0.0135
0.0134	2.6816	360	0.0134
0.0147	2.7188	365	0.0135
0.0127	2.7561	370	0.0134
0.0144	2.7933	375	0.0132
0.0166	2.8305	380	0.0131
0.0136	2.8678	385	0.0131
0.0158	2.9050	390	0.0132
0.0118	2.9423	395	0.0131
0.0133	2.9795	400	0.0130
0.0126	3.0168	405	0.0128
0.0121	3.0540	410	0.0128
0.0127	3.0912	415	0.0128
0.0128	3.1285	420	0.0127
0.0121	3.1657	425	0.0127
0.0121	3.2030	430	0.0127
0.0141	3.2402	435	0.0127
0.013	3.2775	440	0.0126
0.0123	3.3147	445	0.0125
0.0153	3.3520	450	0.0126
0.0148	3.3892	455	0.0126
0.0136	3.4264	460	0.0126
0.0175	3.4637	465	0.0125
0.0143	3.5009	470	0.0125
0.0116	3.5382	475	0.0124
0.012	3.5754	480	0.0123
0.0116	3.6127	485	0.0124
0.0127	3.6499	490	0.0125
0.0158	3.6872	495	0.0124
0.0126	3.7244	500	0.0124
0.0138	3.7616	505	0.0124
0.0135	3.7989	510	0.0124
0.0126	3.8361	515	0.0124
0.0138	3.8734	520	0.0123
0.0117	3.9106	525	0.0123
0.0126	3.9479	530	0.0123
0.0132	3.9851	535	0.0123
0.013	4.0223	540	0.0123
0.0131	4.0596	545	0.0122
0.0156	4.0968	550	0.0122
0.0129	4.1341	555	0.0122
0.0128	4.1713	560	0.0122
0.0098	4.2086	565	0.0121
0.0109	4.2458	570	0.0121
0.0128	4.2831	575	0.0121
0.0116	4.3203	580	0.0121
0.0126	4.3575	585	0.0121
0.0118	4.3948	590	0.0121
0.0136	4.4320	595	0.0121
0.0122	4.4693	600	0.0121
0.0142	4.5065	605	0.0120
0.011	4.5438	610	0.0120
0.011	4.5810	615	0.0120
0.011	4.6182	620	0.0120
0.0104	4.6555	625	0.0120
0.0113	4.6927	630	0.0120
0.0123	4.7300	635	0.0120
0.011	4.7672	640	0.0120
0.0125	4.8045	645	0.0120
0.0121	4.8417	650	0.0120
0.0124	4.8790	655	0.0120
0.0102	4.9162	660	0.0120
0.0151	4.9534	665	0.0121
0.0125	4.9907	670	0.0120

Framework versions

• PEFT 0.12.0
• Transformers 4.46.1
• Pytorch 2.4.0+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3