first commit

app.py

@@ -0,0 +1,211 @@
+import gradio as gr
+import torch
+import transformers
+import transformers
+import torch
+import torch.nn.functional as F
+from torch import nn
+from torch.cuda.amp import custom_fwd, custom_bwd
+from bitsandbytes.functional import quantize_blockwise, dequantize_blockwise
+
+
+class FrozenBNBLinear(nn.Module):
+    def __init__(self, weight, absmax, code, bias=None):
+        assert isinstance(bias, nn.Parameter) or bias is None
+        super().__init__()
+        self.out_features, self.in_features = weight.shape
+        self.register_buffer("weight", weight.requires_grad_(False))
+        self.register_buffer("absmax", absmax.requires_grad_(False))
+        self.register_buffer("code", code.requires_grad_(False))
+        self.adapter = None
+        self.bias = bias
+ 
+    def forward(self, input):
+        output = DequantizeAndLinear.apply(input, self.weight, self.absmax, self.code, self.bias)
+        if self.adapter:
+            output += self.adapter(input)
+        return output
+ 
+    @classmethod
+    def from_linear(cls, linear: nn.Linear) -> "FrozenBNBLinear":
+        weights_int8, state = quantize_blockise_lowmemory(linear.weight)
+        return cls(weights_int8, *state, linear.bias)
+ 
+    def __repr__(self):
+        return f"{self.__class__.__name__}({self.in_features}, {self.out_features})"
+ 
+ 
+class DequantizeAndLinear(torch.autograd.Function): 
+    @staticmethod
+    @custom_fwd
+    def forward(ctx, input: torch.Tensor, weights_quantized: torch.ByteTensor,
+                absmax: torch.FloatTensor, code: torch.FloatTensor, bias: torch.FloatTensor):
+        weights_deq = dequantize_blockwise(weights_quantized, absmax=absmax, code=code)
+        ctx.save_for_backward(input, weights_quantized, absmax, code)
+        ctx._has_bias = bias is not None
+        return F.linear(input, weights_deq, bias)
+ 
+    @staticmethod
+    @custom_bwd
+    def backward(ctx, grad_output: torch.Tensor):
+        assert not ctx.needs_input_grad[1] and not ctx.needs_input_grad[2] and not ctx.needs_input_grad[3]
+        input, weights_quantized, absmax, code = ctx.saved_tensors
+        # grad_output: [*batch, out_features]
+        weights_deq = dequantize_blockwise(weights_quantized, absmax=absmax, code=code)
+        grad_input = grad_output @ weights_deq
+        grad_bias = grad_output.flatten(0, -2).sum(dim=0) if ctx._has_bias else None
+        return grad_input, None, None, None, grad_bias
+ 
+ 
+class FrozenBNBEmbedding(nn.Module):
+    def __init__(self, weight, absmax, code):
+        super().__init__()
+        self.num_embeddings, self.embedding_dim = weight.shape
+        self.register_buffer("weight", weight.requires_grad_(False))
+        self.register_buffer("absmax", absmax.requires_grad_(False))
+        self.register_buffer("code", code.requires_grad_(False))
+        self.adapter = None
+ 
+    def forward(self, input, **kwargs):
+        with torch.no_grad():
+            # note: both quantuized weights and input indices are *not* differentiable
+            weight_deq = dequantize_blockwise(self.weight, absmax=self.absmax, code=self.code)
+            output = F.embedding(input, weight_deq, **kwargs)
+        if self.adapter:
+            output += self.adapter(input)
+        return output 
+ 
+    @classmethod
+    def from_embedding(cls, embedding: nn.Embedding) -> "FrozenBNBEmbedding":
+        weights_int8, state = quantize_blockise_lowmemory(embedding.weight)
+        return cls(weights_int8, *state)
+ 
+    def __repr__(self):
+        return f"{self.__class__.__name__}({self.num_embeddings}, {self.embedding_dim})"
+ 
+ 
+def quantize_blockise_lowmemory(matrix: torch.Tensor, chunk_size: int = 2 ** 20):
+    assert chunk_size % 4096 == 0
+    code = None
+    chunks = []
+    absmaxes = []
+    flat_tensor = matrix.view(-1)
+    for i in range((matrix.numel() - 1) // chunk_size + 1):
+        input_chunk = flat_tensor[i * chunk_size: (i + 1) * chunk_size].clone()
+        quantized_chunk, (absmax_chunk, code) = quantize_blockwise(input_chunk, code=code)
+        chunks.append(quantized_chunk)
+        absmaxes.append(absmax_chunk)
+ 
+    matrix_i8 = torch.cat(chunks).reshape_as(matrix)
+    absmax = torch.cat(absmaxes)
+    return matrix_i8, (absmax, code)
+ 
+ 
+def convert_to_int8(model):
+    """Convert linear and embedding modules to 8-bit with optional adapters"""
+    for module in list(model.modules()):
+        for name, child in module.named_children():
+            if isinstance(child, nn.Linear):
+                print(name, child)
+                setattr(
+                    module,
+                    name,
+                    FrozenBNBLinear(
+                        weight=torch.zeros(child.out_features, child.in_features, dtype=torch.uint8),
+                        absmax=torch.zeros((child.weight.numel() - 1) // 4096 + 1),
+                        code=torch.zeros(256),
+                        bias=child.bias,
+                    ),
+                )
+            elif isinstance(child, nn.Embedding):
+                setattr(
+                    module,
+                    name,
+                    FrozenBNBEmbedding(
+                        weight=torch.zeros(child.num_embeddings, child.embedding_dim, dtype=torch.uint8),
+                        absmax=torch.zeros((child.weight.numel() - 1) // 4096 + 1),
+                        code=torch.zeros(256),
+                    )
+                )
+
+class GPTJBlock(transformers.models.gptj.modeling_gptj.GPTJBlock):
+    def __init__(self, config):
+        super().__init__(config)
+
+        convert_to_int8(self.attn)
+        convert_to_int8(self.mlp)
+
+
+class GPTJModel(transformers.models.gptj.modeling_gptj.GPTJModel):
+    def __init__(self, config):
+        super().__init__(config)
+        convert_to_int8(self)
+        
+
+class GPTJForCausalLM(transformers.models.gptj.modeling_gptj.GPTJForCausalLM):
+    def __init__(self, config):
+        super().__init__(config)
+        convert_to_int8(self)
+
+class T5ForConditionalGeneration(transformers.models.t5.modeling_t5.T5ForConditionalGeneration):
+    def __init__(self, config):
+        super().__init__(config)
+        convert_to_int8(self)
+
+transformers.models.gptj.modeling_gptj.GPTJBlock = GPTJBlock
+transformers.models.t5.modeling_t5.T5ForConditionalGeneration = T5ForConditionalGeneration
+
+config = transformers.GPTJConfig.from_pretrained("EleutherAI/gpt-j-6B")
+tokenizer = transformers.AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")
+
+config.pad_token_id = config.eos_token_id
+tokenizer.pad_token = config.pad_token_id
+
+gpt = GPTJForCausalLM(config)#.from_pretrained("hivemind/gpt-j-6B-8bit", low_cpu_mem_usage=True)
+
+def add_adapters(model, adapter_dim=4, p = 0.1):
+    assert adapter_dim > 0
+
+    for name, module in model.named_modules():
+      if isinstance(module, FrozenBNBLinear):
+          if "attn" in name or "mlp" in name or "head" in name:
+              print("Adding adapter to", name)
+              module.adapter = nn.Sequential(
+                nn.Linear(module.in_features, adapter_dim, bias=False),
+                nn.Dropout(p=p),
+                nn.Linear(adapter_dim, module.out_features, bias=False),
+            )
+              print("Initializing", name)
+              nn.init.zeros_(module.adapter[2].weight)
+
+          else:
+              print("Not adding adapter to", name)
+      elif isinstance(module, FrozenBNBEmbedding):
+          print("Adding adapter to", name)
+          module.adapter = nn.Sequential(
+                nn.Embedding(module.num_embeddings, adapter_dim),
+                nn.Dropout(p=p),
+                nn.Linear(adapter_dim, module.embedding_dim, bias=False),
+            )
+          print("Initializing", name)
+          nn.init.zeros_(module.adapter[2].weight)
+
+add_adapters(gpt)
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+gpt.to(device)
+if device == 'cpu':
+    gpt.load_state_dict(torch.load('rewrite_and_paraphrase_pretrained_gptj8bit.pt', map_location=torch.device('cpu')))
+else:
+    gpt.load_state_dict(torch.load('rewrite_and_paraphrase_pretrained_gptj8bit.pt'))
+gpt.eval()
+
+def inference(text):
+    with torch.no_grad():
+        prompt = tokenizer(text, truncation=True, padding=True, max_length=128, return_tensors='pt')
+        prompt = {key: value for key, value in prompt.items()}
+        out = gpt.generate(**prompt, max_length=512, top_k=50, top_p=0.9, temperature=1.0, do_sample=True, repetition_penalty = 1.2, num_beams=1)
+        return tokenizer.decode(out[0])
+
+
+iface = gr.Interface(fn=inference, inputs="text", outputs="text")
+iface.launch()

requirements.txt

@@ -0,0 +1,5 @@
+torch
+gradio
+transformers==4.14.1
+bitsandbytes-cuda111==0.26.0
+datasets==1.16.1

rewrite_and_paraphrase_pretrained_gptj8bit.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03b1fa07169c705ef7e9f68eccf4e3f7050a8bd12316d754fd5e25f0fb351f9b
+size 6231829230