Merge remote-tracking branch 'thu/main'

config.json

@@ -10,6 +10,7 @@
   },
   "bos_token_id": 130004,
   "eos_token_id": 130005,
+  "pad_token_id": 3,
   "hidden_size": 4096,
   "inner_hidden_size": 16384,
   "layernorm_epsilon": 1e-05,

configuration_chatglm.py

@@ -72,6 +72,8 @@ class ChatGLMConfig(PretrainedConfig):
             position_encoding_2d=True,
             quantization_bit=0,
             quantization_embeddings=False,
+            pre_seq_len=None,
+            prefix_projection=False,
             **kwargs
     ):
         self.num_layers = num_layers
@@ -86,8 +88,11 @@ class ChatGLMConfig(PretrainedConfig):
         self.eos_token_id = eos_token_id
         self.pad_token_id = pad_token_id
         self.position_encoding_2d = position_encoding_2d
-        self.quantization_bit=quantization_bit
-        self.quantization_embeddings=quantization_embeddings
+        self.quantization_bit = quantization_bit
+        self.quantization_embeddings = quantization_embeddings
+        self.pre_seq_len = pre_seq_len
+        self.prefix_projection = prefix_projection
+
         super().__init__(
             pad_token_id=pad_token_id,
             bos_token_id=bos_token_id,

modeling_chatglm.py

@@ -5,6 +5,7 @@ import copy
 import os
 import warnings
 import re
+import sys
 
 import torch
 import torch.utils.checkpoint
@@ -12,7 +13,7 @@ import torch.nn.functional as F
 from torch import nn
 from torch.nn import CrossEntropyLoss, LayerNorm
 from torch.nn.utils import skip_init
-from typing import Optional, Tuple, Union, List, Callable
+from typing import Optional, Tuple, Union, List, Callable, Dict, Any
 
 from transformers.utils import (
     add_code_sample_docstrings,
@@ -27,16 +28,18 @@ from transformers.modeling_outputs import (
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import logging
 from transformers.generation.logits_process import LogitsProcessor
-from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList, GenerationConfig
+from transformers.generation.utils import LogitsProcessorList, StoppingCriteriaList, GenerationConfig, ModelOutput
 
 from .configuration_chatglm import ChatGLMConfig
 
 
 # flags required to enable jit fusion kernels
-torch._C._jit_set_profiling_mode(False)
-torch._C._jit_set_profiling_executor(False)
-torch._C._jit_override_can_fuse_on_cpu(True)
-torch._C._jit_override_can_fuse_on_gpu(True)
+
+if sys.platform != 'darwin':
+    torch._C._jit_set_profiling_mode(False)
+    torch._C._jit_set_profiling_executor(False)
+    torch._C._jit_override_can_fuse_on_cpu(True)
+    torch._C._jit_override_can_fuse_on_gpu(True)
 
 logger = logging.get_logger(__name__)
 
@@ -131,6 +134,36 @@ def load_tf_weights_in_chatglm_6b(model, config, tf_checkpoint_path):
     return model
 
 
+class PrefixEncoder(torch.nn.Module):
+    """
+    The torch.nn model to encode the prefix
+    Input shape: (batch-size, prefix-length)
+    Output shape: (batch-size, prefix-length, 2*layers*hidden)
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.prefix_projection = config.prefix_projection
+        if self.prefix_projection:
+            # Use a two-layer MLP to encode the prefix
+            self.embedding = torch.nn.Embedding(config.pre_seq_len, config.hidden_size)
+            self.trans = torch.nn.Sequential(
+                torch.nn.Linear(config.hidden_size, config.hidden_size),
+                torch.nn.Tanh(),
+                torch.nn.Linear(config.hidden_size, config.num_layers * config.hidden_size * 2)
+            )
+        else:
+            self.embedding = torch.nn.Embedding(config.pre_seq_len, config.num_layers * config.hidden_size * 2)
+
+    def forward(self, prefix: torch.Tensor):
+        if self.prefix_projection:
+            prefix_tokens = self.embedding(prefix)
+            past_key_values = self.trans(prefix_tokens)
+        else:
+            past_key_values = self.embedding(prefix)
+        return past_key_values
+
+
 @torch.jit.script
 def gelu_impl(x):
     """OpenAI's gelu implementation."""
@@ -219,7 +252,7 @@ def attention_fn(
         use_cache=False,
 ):
     if layer_past is not None:
-        past_key, past_value = layer_past
+        past_key, past_value = layer_past[0], layer_past[1]
         key_layer = torch.cat((past_key, key_layer), dim=0)
         value_layer = torch.cat((past_value, value_layer), dim=0)
 
@@ -273,7 +306,7 @@ def attention_fn(
         if not (attention_mask == 0).all():
             # if auto-regressive, skip
             attention_scores.masked_fill_(attention_mask, -10000.0)
-        dtype = attention_scores.type()
+        dtype = attention_scores.dtype
         attention_scores = attention_scores.float()
         attention_scores = attention_scores * query_key_layer_scaling_coeff
 
@@ -619,10 +652,10 @@ class ChatGLMPreTrainedModel(PreTrainedModel):
     """
 
     is_parallelizable = False
-    supports_gradient_checkpointing = False
+    supports_gradient_checkpointing = True
     config_class = ChatGLMConfig
     base_model_prefix = "transformer"
-    _no_split_modules = ["GLM6BBlock"]
+    _no_split_modules = ["GLMBlock"]
 
     def __init__(self, *inputs, **kwargs):
         super().__init__(*inputs, **kwargs)
@@ -631,6 +664,43 @@ class ChatGLMPreTrainedModel(PreTrainedModel):
         """Initialize the weights."""
         return
 
+    def get_masks(self, input_ids, device):
+        batch_size, seq_length = input_ids.shape
+        context_lengths = [seq.tolist().index(self.config.bos_token_id) for seq in input_ids]
+        attention_mask = torch.ones((batch_size, seq_length, seq_length), device=device)
+        attention_mask.tril_()
+        for i, context_length in enumerate(context_lengths):
+            attention_mask[i, :, :context_length] = 1
+        attention_mask.unsqueeze_(1)
+        attention_mask = (attention_mask < 0.5).bool()
+
+        return attention_mask
+
+    def get_position_ids(self, input_ids, mask_positions, device, gmask=False):
+        batch_size, seq_length = input_ids.shape
+        context_lengths = [seq.tolist().index(self.config.bos_token_id) for seq in input_ids]
+        if self.position_encoding_2d:
+            position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1)
+            for i, context_length in enumerate(context_lengths):
+                position_ids[i, context_length:] = mask_positions[i]
+            block_position_ids = [torch.cat((
+                torch.zeros(context_length, dtype=torch.long, device=device),
+                torch.arange(seq_length - context_length, dtype=torch.long, device=device) + 1
+            )) for context_length in context_lengths]
+            block_position_ids = torch.stack(block_position_ids, dim=0)
+            position_ids = torch.stack((position_ids, block_position_ids), dim=1)
+        else:
+            position_ids = torch.arange(seq_length, dtype=torch.long, device=device).unsqueeze(0).repeat(batch_size, 1)
+            if not gmask:
+                for i, context_length in enumerate(context_lengths):
+                    position_ids[context_length:] = mask_positions[i]
+
+        return position_ids
+
+    def _set_gradient_checkpointing(self, module, value=False):
+        if isinstance(module, ChatGLMModel):
+            module.gradient_checkpointing = value
+
 
 CHATGLM_6B_START_DOCSTRING = r"""
     This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) sub-class.
@@ -727,12 +797,15 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
         self.inner_hidden_size = config.inner_hidden_size
         self.hidden_size_per_attention_head = self.hidden_size // self.num_attention_heads
         self.position_encoding_2d = config.position_encoding_2d
+        self.pre_seq_len = config.pre_seq_len
+        self.prefix_projection = config.prefix_projection
 
         self.word_embeddings = skip_init(
             torch.nn.Embedding,
             num_embeddings=self.vocab_size, embedding_dim=self.hidden_size,
             dtype=self.params_dtype
         )
+        self.gradient_checkpointing = False
 
         def get_layer(layer_id):
             return GLMBlock(
@@ -755,43 +828,38 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
         # Final layer norm before output.
         self.final_layernorm = LayerNorm(self.hidden_size, eps=self.layernorm_epsilon)
 
+        if self.pre_seq_len is not None:
+            for param in self.parameters():
+                param.requires_grad = False
+            self.prefix_tokens = torch.arange(self.pre_seq_len).long()
+            self.prefix_encoder = PrefixEncoder(config)
+            self.dropout = torch.nn.Dropout(0.1)
+
+            # total_params = sum(p.numel() for p in self.parameters())
+            # trainable_params = sum(p.numel() for p in self.parameters() if p.requires_grad)
+            # print("Using p-tuning v2: # trainable_params = {} / {}".format(trainable_params, total_params))
+
     def get_input_embeddings(self):
         return self.word_embeddings
 
     def set_input_embeddings(self, new_embeddings: torch.Tensor):
         self.word_embeddings = new_embeddings
 
-    def get_masks(self, seq, device):
-        context_length = seq.index(self.config.bos_token_id) + 1
-
-        attention_mask = torch.ones((1, len(seq), len(seq)), device=device)
-        attention_mask.tril_()
-        attention_mask[..., :context_length - 1] = 1
-        attention_mask.unsqueeze_(1)
-        attention_mask = (attention_mask < 0.5).bool()
-
-        return attention_mask
-
-    def get_position_ids(self, seq, mask_position, device, gmask=False):
-        context_length = seq.index(self.config.bos_token_id) + 1
-        if self.position_encoding_2d:
-            seq_length = seq.index(self.config.bos_token_id)
-            position_ids = torch.arange(context_length, dtype=torch.long, device=device)
-            if not gmask:
-                position_ids[seq_length:] = mask_position
-            block_position_ids = torch.cat((
-                torch.zeros(seq_length, dtype=torch.long, device=device),
-                torch.arange(context_length - seq_length, dtype=torch.long, device=device) + 1
-            ))
-            position_ids = torch.stack((position_ids, block_position_ids), dim=0)
-        else:
-            position_ids = torch.arange(context_length, dtype=torch.long, device=device)
-            if not gmask:
-                position_ids[context_length - 1:] = mask_position
-
-        position_ids = position_ids.unsqueeze(0)
-
-        return position_ids
+    def get_prompt(self, batch_size, device, dtype=torch.half):
+        prefix_tokens = self.prefix_tokens.unsqueeze(0).expand(batch_size, -1).to(device)
+        past_key_values = self.prefix_encoder(prefix_tokens).type(dtype)
+        past_key_values = past_key_values.view(
+            batch_size,
+            self.pre_seq_len,
+            self.num_layers * 2,
+            self.num_attention_heads,
+            self.hidden_size // self.num_attention_heads
+        )
+        # seq_len, b, nh, hidden_size
+        past_key_values = self.dropout(past_key_values)
+        past_key_values = past_key_values.permute([2, 1, 0, 3, 4]).split(2)
+        # past_key_values = [(v[0], v[1]) for v in past_key_values]
+        return past_key_values
 
     @add_start_docstrings_to_model_forward(CHATGLM_6B_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
     @add_code_sample_docstrings(
@@ -819,6 +887,13 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
         use_cache = use_cache if use_cache is not None else self.config.use_cache
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
         if input_ids is not None and inputs_embeds is not None:
             raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
         elif input_ids is not None:
@@ -828,31 +903,41 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
         else:
             raise ValueError("You have to specify either input_ids or inputs_embeds")
 
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
         if past_key_values is None:
-            past_key_values = tuple([None] * len(self.layers))
-            seq = input_ids[0].tolist()
+            if self.pre_seq_len is not None:
+                past_key_values = self.get_prompt(batch_size=input_ids.shape[0], device=input_ids.device,
+                                                  dtype=inputs_embeds.dtype)
+            else:
+                past_key_values = tuple([None] * len(self.layers))
 
             if attention_mask is None:
                 attention_mask = self.get_masks(
-                    seq=seq,
+                    input_ids,
                     device=input_ids.device
                 )
 
+
             if position_ids is None:
                 MASK, gMASK = 130000, 130001
                 mask_token = MASK if MASK in input_ids else gMASK
                 use_gmask = False if MASK in input_ids else gMASK
 
-                mask_position = seq.index(mask_token)
+                mask_positions = [seq.tolist().index(mask_token) for seq in input_ids]
                 position_ids = self.get_position_ids(
-                    seq=seq,
-                    mask_position=mask_position,
+                    input_ids,
+                    mask_positions=mask_positions,
                     device=input_ids.device,
                     gmask=use_gmask
                 )
 
-        if inputs_embeds is None:
-            inputs_embeds = self.word_embeddings(input_ids)
+        if self.pre_seq_len is not None and attention_mask is not None:
+            prefix_attention_mask = torch.ones(batch_size, 1, input_ids.size(-1), self.pre_seq_len).to(
+                attention_mask.device)
+            prefix_attention_mask = (prefix_attention_mask < 0.5).bool()
+            attention_mask = torch.cat((prefix_attention_mask, attention_mask), dim=3)
 
         # [seq_len, batch, hidden_size]
         hidden_states = inputs_embeds.transpose(0, 1)
@@ -861,11 +946,6 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
         all_self_attentions = () if output_attentions else None
         all_hidden_states = () if output_hidden_states else None
 
-        seq_length_with_past = seq_length
-        past_key_values_length = 0
-        if past_key_values[0] is not None:
-            past_key_values_length = past_key_values[0][0].shape[0]
-            seq_length_with_past = seq_length_with_past + past_key_values_length
         if attention_mask is None:
             attention_mask = torch.zeros(1, 1, device=input_ids.device).bool()
 
@@ -876,16 +956,29 @@ class ChatGLMModel(ChatGLMPreTrainedModel):
 
             if output_hidden_states:
                 all_hidden_states = all_hidden_states + (hidden_states,)
-
-            layer_ret = layer(
-                hidden_states,
-                position_ids=position_ids,
-                attention_mask=attention_mask,
-                layer_id=torch.tensor(i),
-                layer_past=past_key_values[i],
-                use_cache=use_cache,
-                output_attentions=output_attentions
-            )
+            layer_past = past_key_values[i]
+
+            if self.gradient_checkpointing and self.training:
+                layer_ret = torch.utils.checkpoint.checkpoint(
+                    layer,
+                    hidden_states,
+                    position_ids,
+                    attention_mask,
+                    torch.tensor(i),
+                    layer_past,
+                    use_cache,
+                    output_attentions
+                )
+            else:
+                layer_ret = layer(
+                    hidden_states,
+                    position_ids=position_ids,
+                    attention_mask=attention_mask,
+                    layer_id=torch.tensor(i),
+                    layer_past=layer_past,
+                    use_cache=use_cache,
+                    output_attentions=output_attentions
+                )
 
             hidden_states = layer_ret[0]
 
@@ -946,31 +1039,40 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
     def set_output_embeddings(self, new_embeddings):
         self.lm_head = new_embeddings
 
-    def get_masks_and_position_ids(self, seq, mask_position, context_length, device, gmask=False):
-        attention_mask = torch.ones((1, context_length, context_length), device=device)
-        attention_mask.tril_()
-        attention_mask[..., :context_length - 1] = 1
-        attention_mask.unsqueeze_(1)
-        attention_mask = (attention_mask < 0.5).bool()
+    def _update_model_kwargs_for_generation(
+        self,
+        outputs: ModelOutput,
+        model_kwargs: Dict[str, Any],
+        is_encoder_decoder: bool = False,
+        standardize_cache_format: bool = False,
+    ) -> Dict[str, Any]:
+        # update past_key_values
+        model_kwargs["past_key_values"] = self._extract_past_from_model_output(
+            outputs, standardize_cache_format=standardize_cache_format
+        )
 
-        if self.position_encoding_2d:
-            seq_length = seq.index(self.config.bos_token_id)
-            position_ids = torch.arange(context_length, dtype=torch.long, device=device)
-            if not gmask:
-                position_ids[seq_length:] = mask_position
-            block_position_ids = torch.cat((
-                torch.zeros(seq_length, dtype=torch.long, device=device),
-                torch.arange(context_length - seq_length, dtype=torch.long, device=device) + 1
-            ))
-            position_ids = torch.stack((position_ids, block_position_ids), dim=0)
-        else:
-            position_ids = torch.arange(context_length, dtype=torch.long, device=device)
-            if not gmask:
-                position_ids[context_length - 1:] = mask_position
+        # update attention mask
+        if "attention_mask" in model_kwargs:
+            attention_mask = model_kwargs["attention_mask"]
+            if attention_mask is not None and attention_mask.dtype == torch.bool:
+                attention_mask = torch.cat(
+                    [attention_mask, attention_mask.new_ones((*attention_mask.shape[:3], 1))], dim=3)
+                new_attention_mask = attention_mask[:, :, -1:].clone()
+                new_attention_mask[..., -1] = False
+                model_kwargs["attention_mask"] = torch.cat(
+                    [attention_mask, new_attention_mask], dim=2
+                )
 
-        position_ids = position_ids.unsqueeze(0)
+        # update position ids
+        if "position_ids" in model_kwargs:
+            position_ids = model_kwargs["position_ids"]
+            new_position_id = position_ids[..., -1:].clone()
+            new_position_id[:, 1, :] += 1
+            model_kwargs["position_ids"] = torch.cat(
+                [position_ids, new_position_id], dim=-1
+            )
 
-        return attention_mask, position_ids
+        return model_kwargs
 
     def prepare_inputs_for_generation(
             self,
@@ -978,27 +1080,34 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
             past: Optional[torch.Tensor] = None,
             past_key_values: Optional[torch.Tensor] = None,
             attention_mask: Optional[torch.Tensor] = None,
+            position_ids: Optional[torch.Tensor] = None,
             **kwargs
     ) -> dict:
-
+        batch_size, seq_length = input_ids.shape
         MASK, gMASK = 130000, 130001
         mask_token = MASK if MASK in input_ids else gMASK
         use_gmask = False if MASK in input_ids else gMASK
-        seq = input_ids[0].tolist()
-        mask_position = seq.index(mask_token)
-
-        if mask_token not in seq:
-            raise ValueError("You have to add either [MASK] or [gMASK] in your input")
+        seqs = input_ids.tolist()
+        mask_positions = [seq.index(mask_token) for seq in seqs]
 
         # only last token for input_ids if past is not None
         if past is not None or past_key_values is not None:
-            context_length = seq.index(self.config.bos_token_id)
             last_token = input_ids[:, -1].unsqueeze(-1)
-            if self.position_encoding_2d:
-                position_ids = torch.tensor([[[mask_position], [len(seq) - context_length]]], dtype=torch.long,
-                                            device=input_ids.device)
+            if attention_mask is not None and attention_mask.dtype == torch.bool:
+                attention_mask = attention_mask[:, :, -1:]
             else:
-                position_ids = torch.tensor([[mask_position]], dtype=torch.long, device=input_ids.device)
+                attention_mask = None
+            if position_ids is not None:
+                position_ids = position_ids[..., -1:]
+            else:
+                context_lengths = [seq.index(self.config.bos_token_id) for seq in seqs]
+                if self.position_encoding_2d:
+                    position_ids = torch.tensor(
+                        [[mask_position, seq_length - context_length] for mask_position, context_length in
+                         zip(mask_positions, context_lengths)], dtype=torch.long, device=input_ids.device).unsqueeze(-1)
+                else:
+                    position_ids = torch.tensor([mask_position for mask_position in mask_positions], dtype=torch.long,
+                                                device=input_ids.device).unsqueeze(-1)
 
             if past is None:
                 past = past_key_values
@@ -1006,15 +1115,24 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
                 "input_ids": last_token,
                 "past_key_values": past,
                 "position_ids": position_ids,
+                "attention_mask": attention_mask
             }
         else:
-            attention_mask, position_ids = self.get_masks_and_position_ids(
-                seq=seq,
-                mask_position=mask_position,
-                context_length=len(seq),
-                device=input_ids.device,
-                gmask=use_gmask
-            )
+            if attention_mask is not None and attention_mask.dtype != torch.bool:
+                logger.warning_once(f"The dtype of attention mask ({attention_mask.dtype}) is not bool")
+                attention_mask = None
+            if attention_mask is None:
+                attention_mask = self.get_masks(
+                    input_ids,
+                    device=input_ids.device
+                )
+            if position_ids is None:
+                position_ids = self.get_position_ids(
+                    input_ids,
+                    device=input_ids.device,
+                    mask_positions=mask_positions,
+                    gmask=use_gmask
+                )
 
             return {
                 "input_ids": input_ids,
@@ -1063,7 +1181,7 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
             shift_logits = lm_logits[..., :-1, :].contiguous()
             shift_labels = labels[..., 1:].contiguous()
             # Flatten the tokens
-            loss_fct = CrossEntropyLoss()
+            loss_fct = CrossEntropyLoss(ignore_index=-100)
             loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))
 
             lm_logits = lm_logits.to(hidden_states.dtype)
@@ -1132,10 +1250,10 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
             for i, (old_query, response) in enumerate(history):
                 prompt += "[Round {}]\n问：{}\n答：{}\n".format(i, old_query, response)
             prompt += "[Round {}]\n问：{}\n答：".format(len(history), query)
-        input_ids = tokenizer([prompt], return_tensors="pt", padding=True)
-        input_ids = input_ids.to(self.device)
-        outputs = self.generate(**input_ids, **gen_kwargs)
-        outputs = outputs.tolist()[0][len(input_ids["input_ids"][0]):]
+        inputs = tokenizer([prompt], return_tensors="pt")
+        inputs = inputs.to(self.device)
+        outputs = self.generate(**inputs, **gen_kwargs)
+        outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):]
         response = tokenizer.decode(outputs)
         response = self.process_response(response)
         history = history + [(query, response)]
@@ -1158,10 +1276,10 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
             for i, (old_query, response) in enumerate(history):
                 prompt += "[Round {}]\n问：{}\n答：{}\n".format(i, old_query, response)
             prompt += "[Round {}]\n问：{}\n答：".format(len(history), query)
-        input_ids = tokenizer([prompt], return_tensors="pt", padding=True)
-        input_ids = input_ids.to(self.device)
-        for outputs in self.stream_generate(**input_ids, **gen_kwargs):
-            outputs = outputs.tolist()[0][len(input_ids["input_ids"][0]):]
+        inputs = tokenizer([prompt], return_tensors="pt")
+        inputs = inputs.to(self.device)
+        for outputs in self.stream_generate(**inputs, **gen_kwargs):
+            outputs = outputs.tolist()[0][len(inputs["input_ids"][0]):]
             response = tokenizer.decode(outputs)
             response = self.process_response(response)
             new_history = history + [(query, response)]
@@ -1298,6 +1416,7 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
                 num_embeddings=self.transformer.word_embeddings.num_embeddings,
                 embedding_dim=self.transformer.word_embeddings.embedding_dim,
                 dtype=torch.half,
+                empty_init=True,
                 device=self.transformer.word_embeddings.weight.device,
             )
             self.lm_head =  QuantizedLinear(
@@ -1310,6 +1429,7 @@ class ChatGLMForConditionalGeneration(ChatGLMPreTrainedModel):
                 quantized_weight=self.transformer.word_embeddings.weight,
                 quantized_weight_scale=self.transformer.word_embeddings.weight_scale,
                 dtype=torch.half,
+                empty_init=True,
                 device=self.lm_head.weight.device,
             )
 

quantization.py

@@ -7,10 +7,13 @@ import bz2
 import torch
 import base64
 import ctypes
+from transformers.utils import logging
 
 from typing import List
 from functools import partial
 
+logger = logging.get_logger(__name__)
+
 try:
     from cpm_kernels.kernels.base import LazyKernelCModule, KernelFunction, round_up
 
@@ -37,18 +40,18 @@ try:
     )
 except Exception as exception:
     kernels = None
-    print("Failed to load cpm_kernels:", exception)
+    logger.warning("Failed to load cpm_kernels:", exception)
 
 
 class W8A16Linear(torch.autograd.Function):
     @staticmethod
     def forward(ctx, inp: torch.Tensor, quant_w: torch.Tensor, scale_w: torch.Tensor, weight_bit_width):
         ctx.inp_shape = inp.size()
-        ctx.weight_shape = quant_w.size()
         ctx.weight_bit_width = weight_bit_width
         out_features = quant_w.size(0)
         inp = inp.contiguous().view(-1, inp.size(-1))
         weight = extract_weight_to_half(quant_w, scale_w, weight_bit_width)
+        ctx.weight_shape = weight.size()
         output = inp.mm(weight.t())
         ctx.save_for_backward(inp, quant_w, scale_w)
         return output.view(*(ctx.inp_shape[:-1] + (out_features,)))
@@ -60,18 +63,18 @@ class W8A16Linear(torch.autograd.Function):
         grad_output = grad_output.contiguous().view(-1, weight.size(0))
         grad_input = grad_output.mm(weight)
         grad_weight = grad_output.t().mm(inp)
-        return grad_input.view(ctx.inp_shape), grad_weight.view(ctx.weight_shape), None
+        return grad_input.view(ctx.inp_shape), grad_weight.view(ctx.weight_shape), None, None
 
 
 class W8A16LinearCPU(torch.autograd.Function):
     @staticmethod
     def forward(ctx, inp: torch.Tensor, quant_w: torch.Tensor, scale_w: torch.Tensor, weight_bit_width, quantization_cache=None):
         ctx.inp_shape = inp.size()
-        ctx.weight_shape = quant_w.size()
         ctx.weight_bit_width = weight_bit_width
         out_features = quant_w.size(0)
         inp = inp.contiguous().view(-1, inp.size(-1))
         weight = extract_weight_to_float(quant_w, scale_w, weight_bit_width, quantization_cache=quantization_cache)
+        ctx.weight_shape = weight.size()
         output = inp.mm(weight.t())
         ctx.save_for_backward(inp, quant_w, scale_w)
         return output.view(*(ctx.inp_shape[:-1] + (out_features,)))
@@ -83,13 +86,13 @@ class W8A16LinearCPU(torch.autograd.Function):
         grad_output = grad_output.contiguous().view(-1, weight.size(0))
         grad_input = grad_output.mm(weight)
         grad_weight = grad_output.t().mm(inp)
-        return grad_input.view(ctx.inp_shape), grad_weight.view(ctx.weight_shape), None
+        return grad_input.view(ctx.inp_shape), grad_weight.view(ctx.weight_shape), None, None
 
 
 default_cpu_kernel_code_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "quantization_kernels.c")
 default_cpu_kernel_code = "QlpoOTFBWSZTWXLbSoQAAgzbgERwQXxmTwAAr/ff3kABt0Q2oRVT0hpo9RtEAAAAyBEiSQ9EGjQGQAAAwANGhowjJoNGmgMEUplMTNSMJ5TQaDJpsoMyRMj8P4mZzFSVVwqSXG8GG7MlVwiToYEQwVD7noBxMhNfkeZYtYFtbgOBUSIGtIQjhNHCEnPJsadhb3yBmRIOD3TeAtNLSaU5GgvKUBWSNuuOIHmVt0YhW6rsmDMDUjeUJGJ64R1Jm5lrh0Aa0tKjhFwPdWcGogxLDSXPWQUWTM8Sd3Qz1HMYNxx3HMeiNqNo4jeRDEfZ3gUSHIcU/heomq0vEzL1Msz5KKGxH8FrNOYw3KaxdqaEmNHYMxJFgQbR0DyRknL2L4kwUSxKRdhjRpEtUqilVfggFL1klaMS3PPRDfNqbBOPWO7m4JTVGhS9QTBDDJaEbLbrUQNB+IpJSKQbG5SZZ5gkwJEhJ3aYKJipZ/i7kinChIOW2lQg"
 default_cpu_parallel_kernel_code_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "quantization_kernels_parallel.c")
-default_cpu_parallel_kernel_code = "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"
+default_cpu_parallel_kernel_code = "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"
 
 cpu_kernels = None
 
@@ -168,7 +171,7 @@ class CPUKernel:
             print("Load kernel :", kernel_file)
         else:
             print("Failed to load kernel.")
-        
+
         if compile_parallel_kernel:
             if parallel_num is None:
                 parallel_num = max(os.cpu_count() // 2, 1)
@@ -176,7 +179,7 @@ class CPUKernel:
             if parallel_num < 4:
                 print("Parallel kernel is not recommended when parallel num < 4.")
             self.SetNumThreads(parallel_num)
-        
+
         self.parallel_num = parallel_num
 
 
@@ -284,10 +287,10 @@ def extract_weight_to_float(weight: torch.Tensor, scale_list: torch.Tensor, sour
 class CacheTensor():
     def __init__(self, *args, **kwargs):
         self.tensor = torch.empty(*args, **kwargs)
-    
+
     def to(self, *args, **kwargs):
         self.tensor = self.tensor.to(*args, **kwargs)
-        
+
     def data_ptr(self):
         return self.tensor.data_ptr()
 
@@ -393,7 +396,7 @@ def load_cpu_kernel(**kwargs):
 
 def quantize(model, weight_bit_width, use_quantization_cache=False, empty_init=False, **kwargs):
     """Replace fp16 linear with quantized linear"""
-    
+
     query_key_value_quantization_cache = None
     dense_quantization_cache = None
     dense_h_to_4h_quantization_cache = None

quantization_kernels_parallel.c

@@ -17,7 +17,7 @@ void compress_int4_weight(void *weight, void *out, int n, int m)
     {
         for(int j=0;j<m;j++)
         {
-            (*(unsigned char*)(out + sizeof(unsigned char) * (i * m + j))) |= ((*(unsigned char*)(weight + sizeof(unsigned char) * (i * (m << 1) + (j << 1)))) << 4);
+            (*(unsigned char*)(out + sizeof(unsigned char) * (i * m + j))) = ((*(unsigned char*)(weight + sizeof(unsigned char) * (i * (m << 1) + (j << 1)))) << 4);
             (*(unsigned char*)(out + sizeof(unsigned char) * (i * m + j))) |= (((*(unsigned char*)(weight + sizeof(unsigned char) * (i * (m << 1) + ((j << 1) | 1)))) & 15));
         }
     }

tokenization_chatglm.py

@@ -1,17 +1,14 @@
 """Tokenization classes for ChatGLM."""
-import sys
-import unicodedata
 from typing import List, Optional, Union
-from functools import lru_cache
 import os
-import collections
-import re
 
 from transformers.tokenization_utils import PreTrainedTokenizer
 from icetk.text_tokenizer import TextTokenizer
-from icetk.utils import auto_create
 import icetk.sentencepiece_model_pb2 as sp_model
-from transformers.utils import logging
+from transformers.utils import logging, PaddingStrategy
+from transformers.tokenization_utils_base import EncodedInput, BatchEncoding
+from typing import Dict
+import numpy as np
 
 logger = logging.get_logger(__name__)
 
@@ -170,7 +167,7 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
 
     vocab_files_names = {"vocab_file": "ice_text.model"}
     max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
-    model_input_names = ["input_ids"]
+    model_input_names = ["input_ids", "attention_mask", "position_ids"]
 
     def __init__(
             self,
@@ -200,7 +197,7 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
         self.eos_token = eos_token
         self.eop_token = eop_token
         self.mask_token = mask_token
-        self.gMASK_token = gmask_token
+        self.gmask_token = gmask_token
 
         self.sp_tokenizer = SPTokenizer(vocab_file)
 
@@ -289,7 +286,7 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
         """
         if os.path.isdir(save_directory):
             vocab_file = os.path.join(
-                save_directory, VOCAB_FILES_NAMES["vocab_file"]
+                save_directory, self.vocab_files_names["vocab_file"]
             )
         else:
             vocab_file = save_directory
@@ -321,10 +318,9 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
         Returns:
             `List[int]`: List of [input IDs](../glossary#input-ids) with the appropriate special tokens.
         """
-        if token_ids_1 is not None:
-            token_ids_0 += token_ids_1
         mask_ids = self.sp_tokenizer[self.mask_token]
-        gmask_ids = self.sp_tokenizer[self.gMASK_token]
+        gmask_ids = self.sp_tokenizer[self.gmask_token]
+        eop_id = self.sp_tokenizer[self.eop_token]
         if mask_ids not in token_ids_0 and gmask_ids not in token_ids_0:
             token_ids_0 += [gmask_ids]
 
@@ -333,4 +329,101 @@ class ChatGLMTokenizer(PreTrainedTokenizer):
 
         token_ids_0 += [self.sp_tokenizer[self.bos_token]]
 
+        if token_ids_1 is not None:
+            if not token_ids_1 or token_ids_1[-1] != eop_id:
+                token_ids_1 += [eop_id]
+            token_ids_0 += token_ids_1
+
         return token_ids_0
+
+    def _pad(
+        self,
+        encoded_inputs: Union[Dict[str, EncodedInput], BatchEncoding],
+        max_length: Optional[int] = None,
+        padding_strategy: PaddingStrategy = PaddingStrategy.DO_NOT_PAD,
+        pad_to_multiple_of: Optional[int] = None,
+        return_attention_mask: Optional[bool] = None,
+    ) -> dict:
+        """
+        Pad encoded inputs (on left/right and up to predefined length or max length in the batch)
+
+        Args:
+            encoded_inputs:
+                Dictionary of tokenized inputs (`List[int]`) or batch of tokenized inputs (`List[List[int]]`).
+            max_length: maximum length of the returned list and optionally padding length (see below).
+                Will truncate by taking into account the special tokens.
+            padding_strategy: PaddingStrategy to use for padding.
+
+                - PaddingStrategy.LONGEST Pad to the longest sequence in the batch
+                - PaddingStrategy.MAX_LENGTH: Pad to the max length (default)
+                - PaddingStrategy.DO_NOT_PAD: Do not pad
+                The tokenizer padding sides are defined in self.padding_side:
+
+                    - 'left': pads on the left of the sequences
+                    - 'right': pads on the right of the sequences
+            pad_to_multiple_of: (optional) Integer if set will pad the sequence to a multiple of the provided value.
+                This is especially useful to enable the use of Tensor Core on NVIDIA hardware with compute capability
+                `>= 7.5` (Volta).
+            return_attention_mask:
+                (optional) Set to False to avoid returning attention mask (default: set to model specifics)
+        """
+        # Load from model defaults
+        bos_token_id = self.sp_tokenizer[self.bos_token]
+        mask_token_id = self.sp_tokenizer[self.mask_token]
+        gmask_token_id = self.sp_tokenizer[self.gmask_token]
+        assert self.padding_side == "left"
+
+        required_input = encoded_inputs[self.model_input_names[0]]
+        seq_length = len(required_input)
+
+        if padding_strategy == PaddingStrategy.LONGEST:
+            max_length = len(required_input)
+
+        if max_length is not None and pad_to_multiple_of is not None and (max_length % pad_to_multiple_of != 0):
+            max_length = ((max_length // pad_to_multiple_of) + 1) * pad_to_multiple_of
+
+        needs_to_be_padded = padding_strategy != PaddingStrategy.DO_NOT_PAD and len(required_input) != max_length
+
+        # Initialize attention mask if not present.
+        if max_length is not None:
+            if "attention_mask" not in encoded_inputs:
+                if bos_token_id in required_input:
+                    context_length = required_input.index(bos_token_id)
+                else:
+                    context_length = seq_length
+                attention_mask = np.ones((1, seq_length, seq_length))
+                attention_mask = np.tril(attention_mask)
+                attention_mask[:, :, :context_length] = 1
+                attention_mask = np.bool_(attention_mask < 0.5)
+                encoded_inputs["attention_mask"] = attention_mask
+
+            if "position_ids" not in encoded_inputs:
+                position_ids = np.arange(seq_length, dtype=np.int64)
+                mask_token = mask_token_id if mask_token_id in required_input else gmask_token_id
+                if mask_token in required_input:
+                    mask_position = required_input.index(mask_token)
+                    position_ids[context_length:] = mask_position
+                block_position_ids = np.concatenate(
+                    [np.zeros(context_length, dtype=np.int64),
+                     np.arange(1, seq_length - context_length + 1, dtype=np.int64)])
+                encoded_inputs["position_ids"] = np.stack([position_ids, block_position_ids], axis=0)
+
+        if needs_to_be_padded:
+            difference = max_length - len(required_input)
+
+            if "attention_mask" in encoded_inputs:
+                encoded_inputs["attention_mask"] = np.pad(encoded_inputs["attention_mask"],
+                                                          pad_width=[(0, 0), (difference, 0), (difference, 0)],
+                                                          mode='constant', constant_values=True)
+            if "token_type_ids" in encoded_inputs:
+                encoded_inputs["token_type_ids"] = [self.pad_token_type_id] * difference + encoded_inputs[
+                    "token_type_ids"
+                ]
+            if "special_tokens_mask" in encoded_inputs:
+                encoded_inputs["special_tokens_mask"] = [1] * difference + encoded_inputs["special_tokens_mask"]
+            if "position_ids" in encoded_inputs:
+                encoded_inputs["position_ids"] = np.pad(encoded_inputs["position_ids"],
+                                                        pad_width=[(0, 0), (difference, 0)])
+            encoded_inputs[self.model_input_names[0]] = [self.pad_token_id] * difference + required_input
+
+        return encoded_inputs