File size: 13,374 Bytes
71289bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import PreTrainedModel, PretrainedConfig, AutoTokenizer, Trainer, TrainingArguments, DataCollatorForLanguageModeling
from datasets import load_dataset
from huggingface_hub import HfApi, create_repo
import math
import os

class ZephyrCoderConfig(PretrainedConfig):
    model_type = "zephyr_coder"
    def __init__(
        self,
        vocab_size=128000,
        hidden_size=2560,
        intermediate_size=10240,
        num_hidden_layers=36,
        num_attention_heads=32,
        num_key_value_heads=8,
        max_position_embeddings=8192,
        rope_theta=1000000.0,
        attention_dropout=0.0,
        hidden_dropout=0.0,
        use_flash_attention=True,
        use_moe=True,
        num_experts=24,
        num_experts_per_tok=6,
        sliding_window_size=4096,
        pad_token_id=0,
        bos_token_id=1,
        eos_token_id=2,
        **kwargs
    ):
        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
        self.vocab_size = vocab_size
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.num_key_value_heads = num_key_value_heads
        self.max_position_embeddings = max_position_embeddings
        self.rope_theta = rope_theta
        self.attention_dropout = attention_dropout
        self.hidden_dropout = hidden_dropout
        self.use_flash_attention = use_flash_attention
        self.use_moe = use_moe
        self.num_experts = num_experts
        self.num_experts_per_tok = num_experts_per_tok
        self.sliding_window_size = sliding_window_size

class RMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.eps = eps
    def forward(self, x):
        variance = x.pow(2).mean(-1, keepdim=True)
        x = x * torch.rsqrt(variance + self.eps)
        return self.weight * x

class RotaryEmbedding(nn.Module):
    def __init__(self, dim, max_position_embeddings=8192, base=1000000.0):
        super().__init__()
        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.base = base
        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
        self.register_buffer("inv_freq", inv_freq)
        self._build_cache(max_position_embeddings)
    def _build_cache(self, seq_len):
        t = torch.arange(seq_len, device=self.inv_freq.device)
        freqs = torch.einsum("i,j->ij", t, self.inv_freq)
        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", emb.cos())
        self.register_buffer("sin_cached", emb.sin())
    def forward(self, x, seq_len=None):
        if seq_len > self.max_position_embeddings:
            self._build_cache(seq_len)
        return self.cos_cached[:seq_len], self.sin_cached[:seq_len]

def rotate_half(x):
    x1, x2 = x.chunk(2, dim=-1)
    return torch.cat((-x2, x1), dim=-1)

def apply_rotary_pos_emb(q, k, cos, sin):
    cos = cos.unsqueeze(0).unsqueeze(0)
    sin = sin.unsqueeze(0).unsqueeze(0)
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed

class GroupedQueryAttention(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.num_kv_heads = config.num_key_value_heads
        self.head_dim = config.hidden_size // config.num_attention_heads
        self.num_groups = self.num_heads // self.num_kv_heads
        self.q_proj = nn.Linear(config.hidden_size, config.hidden_size, bias=False)
        self.k_proj = nn.Linear(config.hidden_size, self.num_kv_heads * self.head_dim, bias=False)
        self.v_proj = nn.Linear(config.hidden_size, self.num_kv_heads * self.head_dim, bias=False)
        self.o_proj = nn.Linear(config.hidden_size, config.hidden_size, bias=False)
        self.dropout = nn.Dropout(config.attention_dropout)
        self.rotary_emb = RotaryEmbedding(self.head_dim, config.max_position_embeddings, config.rope_theta)
    def forward(self, hidden_states, attention_mask=None, position_ids=None, past_key_value=None, use_cache=False, output_attentions=False):
        batch_size, seq_len, _ = hidden_states.shape
        q = self.q_proj(hidden_states).view(batch_size, seq_len, self.num_heads, self.head_dim).transpose(1, 2)
        k = self.k_proj(hidden_states).view(batch_size, seq_len, self.num_kv_heads, self.head_dim).transpose(1, 2)
        v = self.v_proj(hidden_states).view(batch_size, seq_len, self.num_kv_heads, self.head_dim).transpose(1, 2)
        cos, sin = self.rotary_emb(q, seq_len=seq_len)
        q, k = apply_rotary_pos_emb(q, k, cos, sin)
        k = k.repeat_interleave(self.num_groups, dim=1)
        v = v.repeat_interleave(self.num_groups, dim=1)
        attn_weights = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.head_dim)
        if attention_mask is not None:
            attn_weights = attn_weights + attention_mask
        attn_weights = F.softmax(attn_weights, dim=-1, dtype=torch.float32).to(q.dtype)
        attn_weights = self.dropout(attn_weights)
        attn_output = torch.matmul(attn_weights, v)
        attn_output = attn_output.transpose(1, 2).contiguous().reshape(batch_size, seq_len, self.hidden_size)
        attn_output = self.o_proj(attn_output)
        return attn_output, attn_weights

class MoE(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.num_experts = config.num_experts
        self.num_experts_per_tok = config.num_experts_per_tok
        self.gate = nn.Linear(config.hidden_size, config.num_experts, bias=False)
        self.experts = nn.ModuleList([nn.Sequential(
            nn.Linear(config.hidden_size, config.intermediate_size, bias=False),
            nn.GELU(),
            nn.Linear(config.intermediate_size, config.hidden_size, bias=False)
        ) for _ in range(config.num_experts)])
    def forward(self, x):
        batch_size, seq_len, hidden_size = x.shape
        x_flat = x.view(-1, hidden_size)
        gate_logits = self.gate(x_flat)
        gate_weights = F.softmax(gate_logits, dim=-1)
        top_weights, top_indices = torch.topk(gate_weights, self.num_experts_per_tok, dim=-1)
        top_weights = top_weights / top_weights.sum(dim=-1, keepdim=True)
        final_output = torch.zeros_like(x_flat)
        for i in range(self.num_experts):
            mask = (top_indices == i).any(dim=-1)
            if mask.any():
                expert_output = self.experts[i](x_flat[mask])
                weight_mask = (top_indices == i).float()
                weights = (top_weights * weight_mask).sum(dim=-1)
                final_output[mask] += expert_output * weights[mask].unsqueeze(-1)
        return final_output.view(batch_size, seq_len, hidden_size)

class ZephyrCoderBlock(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.self_attn = GroupedQueryAttention(config)
        self.input_layernorm = RMSNorm(config.hidden_size)
        self.mlp = MoE(config) if config.use_moe else nn.Sequential(
            nn.Linear(config.hidden_size, config.intermediate_size, bias=False),
            nn.GELU(),
            nn.Linear(config.intermediate_size, config.hidden_size, bias=False)
        )
        self.post_attention_layernorm = RMSNorm(config.hidden_size)
    def forward(self, hidden_states, attention_mask=None, position_ids=None):
        residual = hidden_states
        hidden_states = self.input_layernorm(hidden_states)
        attn_output, _ = self.self_attn(hidden_states, attention_mask, position_ids)
        hidden_states = residual + attn_output
        residual = hidden_states
        hidden_states = self.post_attention_layernorm(hidden_states)
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + hidden_states
        return hidden_states

class ZephyrCoderModel(PreTrainedModel):
    config_class = ZephyrCoderConfig
    def __init__(self, config):
        super().__init__(config)
        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
        self.layers = nn.ModuleList([ZephyrCoderBlock(config) for _ in range(config.num_hidden_layers)])
        self.norm = RMSNorm(config.hidden_size)
    def forward(self, input_ids=None, attention_mask=None, position_ids=None):
        hidden_states = self.embed_tokens(input_ids)
        if attention_mask is not None:
            attention_mask = attention_mask[:, None, None, :]
            attention_mask = (1.0 - attention_mask) * torch.finfo(hidden_states.dtype).min
        for layer in self.layers:
            hidden_states = layer(hidden_states, attention_mask, position_ids)
        hidden_states = self.norm(hidden_states)
        return hidden_states

class ZephyrCoderForCausalLM(PreTrainedModel):
    config_class = ZephyrCoderConfig
    def __init__(self, config):
        super().__init__(config)
        self.model = ZephyrCoderModel(config)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
    def forward(self, input_ids=None, attention_mask=None, labels=None):
        hidden_states = self.model(input_ids, attention_mask)
        logits = self.lm_head(hidden_states)
        loss = None
        if labels is not None:
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            loss = F.cross_entropy(shift_logits.view(-1, self.config.vocab_size), shift_labels.view(-1))
        return loss, logits
    def generate(self, input_ids, max_length=2048, temperature=0.7, top_p=0.9):
        self.eval()
        with torch.no_grad():
            for _ in range(max_length - input_ids.shape[1]):
                _, logits = self.forward(input_ids=input_ids)
                logits = logits[:, -1, :] / temperature
                sorted_logits, sorted_indices = torch.sort(logits, descending=True)
                cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
                sorted_indices_to_remove = cumulative_probs > top_p
                sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
                sorted_indices_to_remove[..., 0] = 0
                indices_to_remove = sorted_indices_to_remove.scatter(-1, sorted_indices, sorted_indices_to_remove)
                logits[indices_to_remove] = float('-inf')
                probs = F.softmax(logits, dim=-1)
                next_token = torch.multinomial(probs, num_samples=1)
                input_ids = torch.cat([input_ids, next_token], dim=-1)
                if next_token.item() == self.config.eos_token_id:
                    break
        return input_ids

def train_zephyr_coder():
    config = ZephyrCoderConfig()
    model = ZephyrCoderForCausalLM(config)
    tokenizer = AutoTokenizer.from_pretrained("bigcode/starcoder2-15b")
    tokenizer.add_special_tokens({'pad_token': '[PAD]'})
    
    dataset = load_dataset("bigcode/the-stack-dedup", data_dir="data/python", split="train", streaming=True)
    def tokenize_function(examples):
        return tokenizer(examples['content'], truncation=True, max_length=2048, padding=False)
    
    tokenized_dataset = dataset.map(tokenize_function, batched=True, remove_columns=dataset.column_names)
    
    training_args = TrainingArguments(
        output_dir="./zephyr-coder-final",
        num_train_epochs=3,
        per_device_train_batch_size=2,
        gradient_accumulation_steps=16,
        learning_rate=3e-4,
        warmup_steps=2000,
        logging_steps=10,
        save_steps=1000,
        fp16=True,
        gradient_checkpointing=True,
        optim="adamw_8bit",
    )
    
    trainer = Trainer(
        model=model,
        args=training_args,
        train_dataset=tokenized_dataset,
        data_collator=DataCollatorForLanguageModeling(tokenizer, mlm=False),
    )
    
    trainer.train()
    trainer.save_model("./zephyr-coder-final")
    tokenizer.save_pretrained("./zephyr-coder-final")
    return model, tokenizer

def upload_to_huggingface(model_dir="./zephyr-coder-final", repo_name="zephyr-coder-15b"):
    create_repo(repo_name, exist_ok=True)
    api = HfApi()
    api.upload_folder(folder_path=model_dir, repo_id=repo_name)
    print(f"Uploaded to https://huggingface.co/{repo_name}")

def demo():
    tokenizer = AutoTokenizer.from_pretrained("bigcode/starcoder2-15b")
    config = ZephyrCoderConfig()
    model = ZephyrCoderForCausalLM(config)
    
    prompts = [
        "def quicksort(arr):",
        "class TransformerBlock:",
        "def train_neural_network():",
        "async def process_api_request():",
        "def optimize_python_code():",
    ]
    
    for prompt in prompts:
        inputs = tokenizer(prompt, return_tensors="pt")
        outputs = model.generate(inputs.input_ids, max_length=500, temperature=0.7, top_p=0.95)
        print(f"\nPrompt: {prompt}\nGenerated:\n{tokenizer.decode(outputs[0])}\n{'-'*80}")

if __name__ == "__main__":
    model, tokenizer = train_zephyr_coder()
    upload_to_huggingface()
    demo()