初始化项目，由ModelHub XC社区提供模型

Model: rbelanec/train_sst2_42_1779354537
Source: Original Platform

.gitattributes

@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,81 @@
+---
+library_name: transformers
+license: llama3.2
+base_model: meta-llama/Llama-3.2-1B-Instruct
+tags:
+- peft-factory
+- freeze
+- llama-factory
+- generated_from_trainer
+model-index:
+- name: train_sst2_42_1779354537
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# train_sst2_42_1779354537
+
+This model is a fine-tuned version of [meta-llama/Llama-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct) on the sst2 dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0908
+- Num Input Tokens Seen: 3725120
+
+## 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: 2e-06
+- train_batch_size: 8
+- eval_batch_size: 8
+- seed: 42
+- 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: 1
+
+### Training results
+
+| Training Loss | Epoch  | Step | Validation Loss | Input Tokens Seen |
+|:-------------:|:------:|:----:|:---------------:|:-----------------:|
+| 0.0584        | 0.0500 | 379  | 0.1753          | 187072            |
+| 0.1154        | 0.1000 | 758  | 0.1295          | 373504            |
+| 0.0745        | 0.1501 | 1137 | 0.1949          | 557824            |
+| 0.1712        | 0.2001 | 1516 | 0.1069          | 743424            |
+| 0.2865        | 0.2501 | 1895 | 0.1277          | 930944            |
+| 0.1225        | 0.3001 | 2274 | 0.1098          | 1116800           |
+| 0.1152        | 0.3501 | 2653 | 0.1235          | 1303872           |
+| 0.1615        | 0.4002 | 3032 | 0.1323          | 1490688           |
+| 0.0698        | 0.4502 | 3411 | 0.1182          | 1678208           |
+| 0.3465        | 0.5002 | 3790 | 0.1325          | 1864128           |
+| 0.1538        | 0.5502 | 4169 | 0.0976          | 2047552           |
+| 0.1911        | 0.6002 | 4548 | 0.1150          | 2232448           |
+| 0.1499        | 0.6503 | 4927 | 0.0984          | 2420096           |
+| 0.2014        | 0.7003 | 5306 | 0.0908          | 2605504           |
+| 0.0014        | 0.7503 | 5685 | 0.0957          | 2790656           |
+| 0.1294        | 0.8003 | 6064 | 0.0955          | 2979456           |
+| 0.1202        | 0.8503 | 6443 | 0.0970          | 3167488           |
+| 0.0013        | 0.9004 | 6822 | 0.0957          | 3355520           |
+| 0.05          | 0.9504 | 7201 | 0.0956          | 3541632           |
+
+
+### Framework versions
+
+- Transformers 4.51.3
+- Pytorch 2.10.0+cu128
+- Datasets 4.0.0
+- Tokenizers 0.21.4

all_results.json

@@ -0,0 +1,13 @@
+{
+    "epoch": 1.0,
+    "eval_loss": 0.09084735810756683,
+    "eval_runtime": 7.6282,
+    "eval_samples_per_second": 882.909,
+    "eval_steps_per_second": 110.38,
+    "num_input_tokens_seen": 3725120,
+    "total_flos": 2.175051626840064e+16,
+    "train_loss": 0.12470523826255549,
+    "train_runtime": 1215.5483,
+    "train_samples_per_second": 49.866,
+    "train_steps_per_second": 6.233
+}

config.json

@@ -0,0 +1,39 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "head_dim": 64,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 8192,
+  "max_position_embeddings": 131072,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 16,
+  "num_key_value_heads": 8,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "factor": 32.0,
+    "high_freq_factor": 4.0,
+    "low_freq_factor": 1.0,
+    "original_max_position_embeddings": 8192,
+    "rope_type": "llama3"
+  },
+  "rope_theta": 500000.0,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": false,
+  "vocab_size": 128256
+}

eval_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "eval_loss": 0.09084735810756683,
+    "eval_runtime": 7.6282,
+    "eval_samples_per_second": 882.909,
+    "eval_steps_per_second": 110.38,
+    "num_input_tokens_seen": 3725120
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": [
+    128001,
+    128008,
+    128009
+  ],
+  "temperature": 0.6,
+  "top_p": 0.9,
+  "transformers_version": "4.51.3"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58ceb3ca5faec61a359c8f422f68c237c099b498be223e4823117c20a19a3a5c
+size 4417933576

special_tokens_map.json

@@ -0,0 +1,26 @@
+{
+  "additional_special_tokens": [
+    {
+      "content": "<|eom_id|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    }
+  ],
+  "bos_token": {
+    "content": "<|begin_of_text|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|eot_id|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|eot_id|>"
+}

train.yaml

@@ -0,0 +1,65 @@
+seed: 42
+
+### model
+model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
+trust_remote_code: true
+flash_attn: auto
+use_cache: false
+
+### method
+# Full fine-tune of every decoder block, but with the (tied) embeddings frozen.
+# `finetuning_type: freeze` only trains modules whose name matches a trainable layer;
+# embed_tokens / lm_head / final model.norm are "extra" modules and stay frozen unless
+# listed in freeze_extra_modules. Setting freeze_trainable_layers = num_hidden_layers (16
+# for Llama-3.2-1B) makes ALL decoder blocks trainable, so this == "full FT minus
+# embeddings". Because tie_word_embeddings=true, freezing embed_tokens also freezes lm_head.
+# This is lever B of the embedding-amplification fix (see figures/amplification/README.md).
+stage: sft
+do_train: true
+finetuning_type: freeze
+freeze_trainable_layers: 16
+freeze_trainable_modules: all
+# freeze_extra_modules: left unset -> embed_tokens, lm_head (tied), final norm stay frozen
+
+### dataset
+dataset: sst2
+template: llama3
+cutoff_len: 2048
+overwrite_cache: true
+preprocessing_num_workers: 4
+dataloader_num_workers: 4
+packing: false
+
+### output
+output_dir: saves_bts_preliminary/freeze/llama-3.2-1b-instruct/train_sst2_42_1779354537
+logging_steps: 5
+save_steps: 0.05
+overwrite_output_dir: true
+save_only_model: false
+plot_loss: true
+include_num_input_tokens_seen: true
+push_to_hub: true
+push_to_hub_organization: rbelanec
+load_best_model_at_end: true
+save_total_limit: 1
+
+### train
+per_device_train_batch_size: 8
+learning_rate: 2.0e-6
+num_train_epochs: 1
+weight_decay: 1.0e-2
+lr_scheduler_type: cosine
+bf16: true
+ddp_timeout: 180000000
+resume_from_checkpoint: null
+warmup_ratio: 0.1
+optim: adamw_torch
+report_to:
+- wandb
+run_name: freeze_llama-3.2-1b-instruct_train_sst2_42_1779354537
+
+### eval
+per_device_eval_batch_size: 8
+eval_strategy: steps
+eval_steps: 0.05
+val_size: 0.1

train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 1.0,
+    "num_input_tokens_seen": 3725120,
+    "total_flos": 2.175051626840064e+16,
+    "train_loss": 0.12470523826255549,
+    "train_runtime": 1215.5483,
+    "train_samples_per_second": 49.866,
+    "train_steps_per_second": 6.233
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be5956f3244a54af871b31e72ef81b8b4797a3bf991a496b7013ecba3258ed7
+size 6289