fix flash-attn, xformers, packing, support chatml

src/axolotl/monkeypatch/llama_attn_hijack_flash.py

@@ -38,6 +38,18 @@ def get_cu_seqlens(attn_mask):
     )
     # Calculate the sequence lengths
     seq_lengths = change_indices[1:] - change_indices[:-1]
+    # Calculate the length of the final sequence or padding
+    final_seq_length = attn_mask.shape[1] - change_indices[-1]
+    # Append the length of the final sequence or padding to seq_lengths
+    if final_seq_length.item():
+        seq_lengths = torch.cat(
+            [
+                seq_lengths,
+                torch.tensor(
+                    [final_seq_length.item()], dtype=torch.int32, device=device
+                ),
+            ]
+        )
     # Calculate the cumulative sequence lengths
     cu_seqlens = torch.cat(
         [torch.tensor([0], dtype=torch.int32, device=device), seq_lengths.cumsum(0)]

src/axolotl/monkeypatch/llama_attn_hijack_xformers.py

@@ -128,7 +128,8 @@ def xformers_forward(
                 query_states,
                 key_states,
                 value_states,
-                attn_bias=xformers.ops.LowerTriangularMask(),
+                attn_bias=attention_mask,
+                # attn_bias=xformers.ops.LowerTriangularMask(),
             )
         attn_weights = None
     else:

src/axolotl/monkeypatch/llama_expand_mask.py

@@ -10,7 +10,8 @@ def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int]
     """
     Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
     This expansion handles packed sequences so that sequences share the same attention mask integer value
-    when they attend to each other within that sequence. This should result in a block diagonal mask
+    when they attend to each other within that sequence.
+    This expansion transforms the mask to lower triangular form to prevent future peeking.
     """
     bsz, src_len = mask.size()
     tgt_len = tgt_len if tgt_len is not None else src_len
@@ -29,9 +30,14 @@ def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int]
     # we multiply by the binary mask so that 0's in the original mask are correctly excluded
     zero_one_mask = torch.eq(mask, mask.transpose(-1, -2)).int() * binary_mask
 
-    # Expand the mask to the correct dimensions for the current batch index
-    expanded_mask = zero_one_mask.expand(bsz, 1, tgt_len, src_len)
-    inverted_mask = 1.0 - expanded_mask
+    # Now let's create a lower triangular mask of ones that will zero out the upper triangular part
+    lower_triangular_ones = torch.tril(torch.ones((tgt_len, src_len), dtype=dtype)).to(
+        mask.device
+    )
+
+    # Use the lower triangular mask to zero out the upper triangular part of the zero_one_mask
+    masked_zero_one_mask = zero_one_mask * lower_triangular_ones
+    inverted_mask = 1.0 - masked_zero_one_mask
 
     return inverted_mask.masked_fill(
         inverted_mask.to(torch.bool), torch.finfo(dtype).min

src/axolotl/prompt_strategies/alpaca_w_system.py

@@ -66,7 +66,11 @@ class SystemDataPrompter(AlpacaPrompter):
     ) -> Generator[str, None, None]:
         # returns the full prompt from instruction and optional input
         # if a label (=response, =output) is provided, it's also appended.
-        formatted_sys_prompt = f"### System:\n{system}\n\n" if system else ""
+        formatted_sys_prompt = (
+            self.system_format.format(system=system)
+            if system and self.system_format
+            else ""
+        )
         if input:
             res = formatted_sys_prompt + self.turn_format.format(
                 instruction=instruction, input=input
@@ -90,8 +94,15 @@ class OpenOrcaSystemDataPrompter(SystemDataPrompter):
             self.turn_format = "### User:\n{instruction}\n\n### Additional Context:\n{input}\n\n### Assistant:\n"
             self.turn_no_input_format = "### User:\n{instruction}\n\n### Assistant:\n"
         if self.prompt_style == PromptStyle.CHAT.value:
-            self.turn_format = "USER: {instruction}\n{input}\nASSISTANT:"
-            self.turn_no_input_format = "USER: {instruction}\nASSISTANT:"
+            self.turn_format = "User: {instruction}\n{input}\nAssistant:"
+            self.turn_no_input_format = "User: {instruction}\nAssistant:"
+            self.system_format = "System: {system}\n"
+        if self.prompt_style == PromptStyle.CHATML.value:
+            self.turn_format = "<|im_start|>user\n{instruction}\n{input}<|im_end|>\n<|im_start|>assistant\n"
+            self.turn_no_input_format = (
+                "<|im_start|>user\n{instruction}<|im_end|>\n<|im_start|>assistant\n"
+            )
+            self.system_format = "<|im_start|>{system}<|im_end|>\n"
 
 
 class OpenOrcaPromptTokenizingStrategy(InstructionWSystemPromptTokenizingStrategy):
@@ -137,3 +148,12 @@ def load_open_orca(tokenizer, cfg):
         cfg.train_on_inputs,
         cfg.sequence_len,
     )
+
+
+def load_open_orca_chatml(tokenizer, cfg):
+    return OpenOrcaPromptTokenizingStrategy(
+        OpenOrcaSystemDataPrompter(PromptStyle.CHATML.value),
+        tokenizer,
+        cfg.train_on_inputs,
+        cfg.sequence_len,
+    )

src/axolotl/prompters.py

@@ -16,6 +16,7 @@ class PromptStyle(Enum):
 
     INSTRUCT = "instruct"
     CHAT = "chat"
+    CHATML = "chatml"
 
 
 class AlpacaPrompter:
@@ -25,6 +26,7 @@ class AlpacaPrompter:
 
     system_prompt = "Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n"
     system_no_input_prompt = "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
+    system_format: str
     turn_format: str
     turn_no_input_format: str
     prompt_style: Optional[PromptStyle] = None
@@ -39,9 +41,11 @@ class AlpacaPrompter:
             self.turn_no_input_format = (
                 "### Instruction:\n{instruction}\n\n### Response:\n"
             )
+            self.system_format = "### System:\n{system}\n\n"
         if self.prompt_style == PromptStyle.CHAT.value:
             self.turn_format = "USER: {instruction}\n{input}\nASSISTANT:"
             self.turn_no_input_format = "USER: {instruction}\nASSISTANT:"
+            self.system_format = "SYSTEM:{system}\n"
 
     def build_prompt(
         self,

src/axolotl/utils/dataloader.py

@@ -259,7 +259,7 @@ class MultipackDistributedDataloader:
             f"packing_efficiency_estimate: {self.packing_efficiency_estimate} "
             f"actual packing efficiency: {self.efficiency()}"
         )
-        return self._len_est()
+        return max(1, self._len_est())
 
     def efficiency(self):
         return self.eff_total_used / self.eff_total_slots

tests/test_expand_mask.py

@@ -20,8 +20,8 @@ class TestExpandMask(unittest.TestCase):
             [
                 [
                     [
-                        [0.0000e00, 0.0000e00, 0.0000e00, -3.4028e38],
-                        [0.0000e00, 0.0000e00, 0.0000e00, -3.4028e38],
+                        [0.0000e00, -3.4028e38, -3.4028e38, -3.4028e38],
+                        [0.0000e00, 0.0000e00, -3.4028e38, -3.4028e38],
                         [0.0000e00, 0.0000e00, 0.0000e00, -3.4028e38],
                         [-3.4028e38, -3.4028e38, -3.4028e38, 0.0000e00],
                     ]
@@ -29,14 +29,14 @@ class TestExpandMask(unittest.TestCase):
                 [
                     [
                         [0.0000e00, -3.4028e38, -3.4028e38, -3.4028e38],
-                        [-3.4028e38, 0.0000e00, 0.0000e00, -3.4028e38],
+                        [-3.4028e38, 0.0000e00, -3.4028e38, -3.4028e38],
                         [-3.4028e38, 0.0000e00, 0.0000e00, -3.4028e38],
                         [-3.4028e38, -3.4028e38, -3.4028e38, -3.4028e38],
                     ]
                 ],
             ]
         )
-
+        print(repr(_expand_mask(mask, dtype)))
         # Check that the output matches the expected output
         self.assertTrue(torch.allclose(_expand_mask(mask, dtype), expected_output))