Merge branch 'main' of https://github.com/OpenAccess-AI-Collective/axolotl into qlora-openllama-3b-example

.github/workflows/base.yml

@@ -1,4 +1,4 @@
-name: ci-cd
+name: ci-cd-base
 
 on:
   push:

docker/Dockerfile-base

@@ -62,6 +62,7 @@ RUN git clone https://github.com/microsoft/DeepSpeed.git && \
 FROM base-builder AS bnb-builder
 
 WORKDIR /workspace
+ENV CUDA_VERSION_BNB=$CUDA_VERSION_BNB
 
 RUN git clone https://github.com/TimDettmers/bitsandbytes.git && \
     cd bitsandbytes && \
@@ -70,6 +71,8 @@ RUN git clone https://github.com/TimDettmers/bitsandbytes.git && \
 
 FROM base-builder
 
+ENV CUDA_VERSION_BNB=$CUDA_VERSION_BNB
+
 # recompile apex
 RUN python3 -m pip uninstall -y apex
 RUN git clone https://github.com/NVIDIA/apex

scripts/finetune.py

@@ -178,6 +178,15 @@ def train(
             tokenizer, cfg, DEFAULT_DATASET_PREPARED_PATH
         )
 
+    if cfg.debug or "debug" in kwargs:
+        logging.info("check_dataset_labels...")
+        check_dataset_labels(
+            train_dataset.select(
+                [random.randrange(0, len(train_dataset) - 1) for i in range(5)]
+            ),
+            tokenizer,
+        )
+
     if prepare_ds_only:
         logging.info("Finished preparing dataset. Exiting...")
         return
@@ -213,15 +222,6 @@ def train(
         model.save_pretrained(cfg.output_dir)
         return
 
-    if cfg.debug:
-        logging.info("check_dataset_labels...")
-        check_dataset_labels(
-            train_dataset.select(
-                [random.randrange(0, len(train_dataset) - 1) for i in range(5)]
-            ),
-            tokenizer,
-        )
-
     trainer = setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer)
 
     model.config.use_cache = False

src/axolotl/prompt_tokenizers.py

@@ -268,6 +268,9 @@ class AlpacaReflectionPTStrategy(ReflectionPromptTokenizingStrategy):
 
 
 class ShareGPTPromptTokenizingStrategy(PromptTokenizingStrategy):
+    def get_conversation_thread(self, prompt):
+        return prompt["conversations"]
+
     def tokenize_prompt(self, prompt):
         result = {
             "input_ids": [],
@@ -279,7 +282,7 @@ class ShareGPTPromptTokenizingStrategy(PromptTokenizingStrategy):
         assistant_token = self._get_assistant_token()
         try:
             for i, part in enumerate(
-                self.prompter.build_prompt(prompt["conversations"])
+                self.prompter.build_prompt(self.get_conversation_thread(prompt))
             ):
                 if isinstance(part, tuple):
                     if part[0] == "USER:":