update colabs

huggingface · Jun 20, 2024 · 73e0b5e · 73e0b5e
1 parent fd6390c
commit 73e0b5e
Show file tree

Hide file tree

Showing 4 changed files with 63 additions and 49 deletions.
diff --git a/colabs/AutoTrain.ipynb b/colabs/AutoTrain.ipynb
@@ -42,7 +42,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.14"
+   "version": "3.1.-1"
   }
  },
  "nbformat": 4,

diff --git a/colabs/AutoTrain_LLM.ipynb b/colabs/AutoTrain_LLM.ipynb
@@ -21,9 +21,12 @@
         "#@markdown - click `Runtime > Run all` or run each cell individually\n",
         "#@markdown - report issues / feature requests here: https://github.com/huggingface/autotrain-advanced/issues\n",
         "\n",
+        "\n",
         "import os\n",
-        "!pip install -U autotrain-advanced > install_logs.txt\n",
-        "!autotrain setup --colab > setup_logs.txt"
+        "!pip install -U autotrain-advanced > install_logs.txt 2>&1\n",
+        "!autotrain setup --colab > setup_logs.txt\n",
+        "from autotrain import __version__\n",
+        "print(f'AutoTrain version: {__version__}')"
       ]
     },
     {
@@ -53,40 +56,64 @@
         "\n",
         "#@markdown ---\n",
         "#@markdown #### Hyperparameters\n",
+        "unsloth = False # @param [\"False\", \"True\"] {type:\"raw\"}\n",
         "learning_rate = 2e-4 # @param {type:\"number\"}\n",
         "num_epochs = 1 #@param {type:\"number\"}\n",
         "batch_size = 1 # @param {type:\"slider\", min:1, max:32, step:1}\n",
         "block_size = 1024 # @param {type:\"number\"}\n",
-        "trainer = \"sft\" # @param [\"default\", \"sft\", \"orpo\"] {type:\"raw\"}\n",
+        "trainer = \"sft\" # @param [\"generic\", \"sft\"] {type:\"string\"}\n",
         "warmup_ratio = 0.1 # @param {type:\"number\"}\n",
         "weight_decay = 0.01 # @param {type:\"number\"}\n",
         "gradient_accumulation = 4 # @param {type:\"number\"}\n",
-        "mixed_precision = \"fp16\" # @param [\"fp16\", \"bf16\", \"none\"] {type:\"raw\"}\n",
+        "mixed_precision = \"fp16\" # @param [\"fp16\", \"bf16\", \"none\"] {type:\"string\"}\n",
         "peft = True # @param [\"False\", \"True\"] {type:\"raw\"}\n",
-        "quantization = \"int4\" # @param [\"int4\", \"int8\", \"none\"] {type:\"raw\"}\n",
+        "quantization = \"int4\" # @param [\"int4\", \"int8\", \"none\"] {type:\"string\"}\n",
         "lora_r = 16 #@param {type:\"number\"}\n",
         "lora_alpha = 32 #@param {type:\"number\"}\n",
         "lora_dropout = 0.05 #@param {type:\"number\"}\n",
         "\n",
-        "os.environ[\"PROJECT_NAME\"] = project_name\n",
-        "os.environ[\"MODEL_NAME\"] = model_name\n",
-        "os.environ[\"PUSH_TO_HUB\"] = str(push_to_hub)\n",
         "os.environ[\"HF_TOKEN\"] = hf_token\n",
-        "os.environ[\"LEARNING_RATE\"] = str(learning_rate)\n",
-        "os.environ[\"NUM_EPOCHS\"] = str(num_epochs)\n",
-        "os.environ[\"BATCH_SIZE\"] = str(batch_size)\n",
-        "os.environ[\"BLOCK_SIZE\"] = str(block_size)\n",
-        "os.environ[\"WARMUP_RATIO\"] = str(warmup_ratio)\n",
-        "os.environ[\"WEIGHT_DECAY\"] = str(weight_decay)\n",
-        "os.environ[\"GRADIENT_ACCUMULATION\"] = str(gradient_accumulation)\n",
-        "os.environ[\"MIXED_PRECISION\"] = str(mixed_precision)\n",
-        "os.environ[\"PEFT\"] = str(peft)\n",
-        "os.environ[\"QUANTIZATION\"] = str(quantization)\n",
-        "os.environ[\"LORA_R\"] = str(lora_r)\n",
-        "os.environ[\"LORA_ALPHA\"] = str(lora_alpha)\n",
-        "os.environ[\"LORA_DROPOUT\"] = str(lora_dropout)\n",
         "os.environ[\"HF_USERNAME\"] = hf_username\n",
-        "os.environ[\"TRAINER\"] = trainer"
+        "\n",
+        "conf = f\"\"\"\n",
+        "task: llm-{trainer}\n",
+        "base_model: {model_name}\n",
+        "project_name: {project_name}\n",
+        "log: tensorboard\n",
+        "backend: local\n",
+        "\n",
+        "data:\n",
+        "  path: data/\n",
+        "  train_split: train\n",
+        "  valid_split: null\n",
+        "  chat_template: null\n",
+        "  column_mapping:\n",
+        "    text_column: text\n",
+        "\n",
+        "params:\n",
+        "  block_size: {block_size}\n",
+        "  lr: {learning_rate}\n",
+        "  warmup_ratio: {warmup_ratio}\n",
+        "  weight_decay: {weight_decay}\n",
+        "  epochs: {num_epochs}\n",
+        "  batch_size: {batch_size}\n",
+        "  gradient_accumulation: {gradient_accumulation}\n",
+        "  mixed_precision: {mixed_precision}\n",
+        "  peft: {peft}\n",
+        "  quantization: {quantization}\n",
+        "  lora_r: {lora_r}\n",
+        "  lora_alpha: {lora_alpha}\n",
+        "  lora_dropout: {lora_dropout}\n",
+        "  unsloth: {unsloth}\n",
+        "\n",
+        "hub:\n",
+        "  username: ${{HF_USERNAME}}\n",
+        "  token: ${{HF_TOKEN}}\n",
+        "  push_to_hub: {push_to_hub}\n",
+        "\"\"\"\n",
+        "\n",
+        "with open(\"conf.yaml\", \"w\") as f:\n",
+        "  f.write(conf)"
       ]
     },
     {
@@ -98,28 +125,7 @@
       },
       "outputs": [],
       "source": [
-        "!autotrain llm \\\n",
-        "--train \\\n",
-        "--model ${MODEL_NAME} \\\n",
-        "--project-name ${PROJECT_NAME} \\\n",
-        "--data-path data/ \\\n",
-        "--text-column text \\\n",
-        "--lr ${LEARNING_RATE} \\\n",
-        "--batch-size ${BATCH_SIZE} \\\n",
-        "--epochs ${NUM_EPOCHS} \\\n",
-        "--block-size ${BLOCK_SIZE} \\\n",
-        "--warmup-ratio ${WARMUP_RATIO} \\\n",
-        "--lora-r ${LORA_R} \\\n",
-        "--lora-alpha ${LORA_ALPHA} \\\n",
-        "--lora-dropout ${LORA_DROPOUT} \\\n",
-        "--weight-decay ${WEIGHT_DECAY} \\\n",
-        "--gradient-accumulation ${GRADIENT_ACCUMULATION} \\\n",
-        "--quantization ${QUANTIZATION} \\\n",
-        "--mixed-precision ${MIXED_PRECISION} \\\n",
-        "--username ${HF_USERNAME} \\\n",
-        "--trainer ${TRAINER} \\\n",
-        "$( [[ \"$PEFT\" == \"True\" ]] && echo \"--peft\" ) \\\n",
-        "$( [[ \"$PUSH_TO_HUB\" == \"True\" ]] && echo \"--push-to-hub --token ${HF_TOKEN}\" )"
+        "!autotrain --config conf.yaml"
       ]
     }
   ],
@@ -134,8 +140,16 @@
       "name": "python3"
     },
     "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
       "name": "python",
-      "version": "3.10.11"
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.10.14"
     }
   },
   "nbformat": 4,

diff --git a/requirements.txt b/requirements.txt
@@ -48,7 +48,7 @@ itsdangerous==2.2.0
 seqeval==1.2.2
 httpx==0.27.0
 pyyaml==6.0.1
-timm==1.0.3
+timm==1.0.7
 torchmetrics==1.4.0
 pycocotools==2.0.7
-sentence-transformers==3.0.0
+sentence-transformers==3.0.1
diff --git a/src/autotrain/__init__.py b/src/autotrain/__init__.py
@@ -41,7 +41,7 @@
 warnings.filterwarnings("ignore", category=UserWarning, module="huggingface_hub")
 
 logger = Logger().get_logger()
-__version__ = "0.7.127"
+__version__ = "0.7.128"
 
 
 def is_colab():