Merge branch 'master' into grad_accum

.gitignore

@@ -0,0 +1,4 @@
+.DS_Store
+.ipynb_checkpoints/
+__pycache__/
+*.pyc

README.md

@@ -19,7 +19,7 @@ Dependencies:
 - `pip install datasets` for huggingface datasets <3 (if you want to download + preprocess OpenWebText)
 - `pip install tiktoken` for OpenAI's fast BPE code <3
 - `pip install wandb` for optional logging <3
-- `pip install tqdm`
+- `pip install tqdm` <3
 
 ## quick start
 
@@ -37,7 +37,7 @@ This creates a `train.bin` and `val.bin` in that data directory. Now it is time
 $ python train.py config/train_shakespeare_char.py
 ```
 
-If you peak inside it, you'll see that we're training a GPT with a context size of up to 256 characters, 384 feature channels, and it is a 6-layer Transformer with 6 heads in each layer. On one A100 GPU this training run takes about 3 minutes and the best validation loss is 1.4697. Based on the configuration, the model checkpoints are being written into the `--out_dir` directory `out-shakespeare-char`. So once the training finishes we can sample from the best model by pointing the sampling script at this directory:
+If you peek inside it, you'll see that we're training a GPT with a context size of up to 256 characters, 384 feature channels, and it is a 6-layer Transformer with 6 heads in each layer. On one A100 GPU this training run takes about 3 minutes and the best validation loss is 1.4697. Based on the configuration, the model checkpoints are being written into the `--out_dir` directory `out-shakespeare-char`. So once the training finishes we can sample from the best model by pointing the sampling script at this directory:
 
 ```
 $ python sample.py --out_dir=out-shakespeare-char
@@ -84,7 +84,7 @@ bot thou the sought bechive in that to doth groan you,
 No relving thee post mose the wear
 ```
 
-Not bad for ~3 minutes on a CPU, for a hint of the right character gestalt. If you're willing to wait longer free to tune the hyperparameters, increase the size of the network, the context length (`--block_size`), the length of training, etc.
+Not bad for ~3 minutes on a CPU, for a hint of the right character gestalt. If you're willing to wait longer, feel free to tune the hyperparameters, increase the size of the network, the context length (`--block_size`), the length of training, etc.
 
 Finally, on Apple Silicon Macbooks and with a recent PyTorch version make sure to add `--device mps` (short for "Metal Performance Shaders"); PyTorch then uses the on-chip GPU that can *significantly* accelerate training (2-3X) and allow you to use larger networks. See [Issue 28](https://github.com/karpathy/nanoGPT/issues/28) for more.
 

data/openwebtext/prepare.py

@@ -54,12 +54,16 @@ for split, dset in tokenized.items():
     filename = os.path.join(os.path.dirname(__file__), f'{split}.bin')
     dtype = np.uint16 # (can do since enc.max_token_value == 50256 is < 2**16)
     arr = np.memmap(filename, dtype=dtype, mode='w+', shape=(arr_len,))
+    total_batches = 1024
 
-    print(f"writing {filename}...")
     idx = 0
-    for example in tqdm(dset):
+    for batch_idx in tqdm(range(total_batches), desc=f'writing {filename}'):
-        arr[idx : idx + example['len']] = example['ids']
+        # Batch together samples for faster write
-        idx += example['len']
+        batch = dset.shard(num_shards=total_batches, index=batch_idx, contiguous=True).with_format('numpy')
+        arr_batch = np.concatenate(batch['ids'])
+        # Write into mmap
+        arr[idx : idx + len(arr_batch)] = arr_batch
+        idx += len(arr_batch)
     arr.flush()
 
 # train.bin is ~17GB, val.bin ~8.5MB

model.py

@@ -61,7 +61,7 @@ class CausalSelfAttention(nn.Module):
         B, T, C = x.size() # batch size, sequence length, embedding dimensionality (n_embd)
 
         # calculate query, key, values for all heads in batch and move head forward to be the batch dim
-        q, k ,v  = self.c_attn(x).split(self.n_embd, dim=2)
+        q, k, v  = self.c_attn(x).split(self.n_embd, dim=2)
         k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
         q = q.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
         v = v.view(B, T, self.n_head, C // self.n_head).transpose(1, 2) # (B, nh, T, hs)
@@ -69,7 +69,7 @@ class CausalSelfAttention(nn.Module):
         # causal self-attention; Self-attend: (B, nh, T, hs) x (B, nh, hs, T) -> (B, nh, T, T)
         if self.flash:
             # efficient attention using Flash Attention CUDA kernels
-            y = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=self.dropout, is_causal=True)
+            y = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=None, dropout_p=self.dropout if self.training else 0, is_causal=True)
         else:
             # manual implementation of attention
             att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(k.size(-1)))
@@ -207,6 +207,7 @@ class GPT(nn.Module):
         self.config.block_size = block_size
         self.transformer.wpe.weight = nn.Parameter(self.transformer.wpe.weight[:block_size])
         for block in self.transformer.h:
+            if hasattr(block.attn, 'bias'):
                 block.attn.bias = block.attn.bias[:,:,:block_size,:block_size]
 
     @classmethod

train.py

@@ -84,6 +84,7 @@ if ddp:
     init_process_group(backend=backend)
     ddp_rank = int(os.environ['RANK'])
     ddp_local_rank = int(os.environ['LOCAL_RANK'])
+    ddp_world_size = int(os.environ['WORLD_SIZE'])
     device = f'cuda:{ddp_local_rank}'
     torch.cuda.set_device(device)
     master_process = ddp_rank == 0 # this process will do logging, checkpointing etc.
@@ -94,6 +95,9 @@ else:
     # if not ddp, we are running on a single gpu, and one process
     master_process = True
     seed_offset = 0
+    ddp_world_size = 1
+tokens_per_iter = gradient_accumulation_steps * ddp_world_size * batch_size * block_size
+print(f"tokens per iteration will be: {tokens_per_iter:,}")
 
 if master_process:
     os.makedirs(out_dir, exist_ok=True)
@@ -190,6 +194,7 @@ scaler = torch.cuda.amp.GradScaler(enabled=(dtype == 'float16'))
 optimizer = model.configure_optimizers(weight_decay, learning_rate, (beta1, beta2), device_type)
 if init_from == 'resume':
     optimizer.load_state_dict(checkpoint['optimizer'])
+checkpoint = None # free up memory
 
 # compile the model
 if compile:
@@ -288,6 +293,7 @@ while True:
             model.require_backward_grad_sync = (micro_step == gradient_accumulation_steps - 1)
         with ctx:
             logits, loss = model(X, Y)
+            loss = loss / gradient_accumulation_steps # scale the loss to account for gradient accumulation
         # immediately async prefetch next batch while model is doing the forward pass on the GPU
         X, Y = get_batch('train')
         # backward pass, with gradient scaling if training in fp16
@@ -307,7 +313,9 @@ while True:
     dt = t1 - t0
     t0 = t1
     if iter_num % log_interval == 0 and master_process:
-        lossf = loss.item() # loss as float. note: this is a CPU-GPU sync point
+        # get loss as float. note: this is a CPU-GPU sync point
+        # scale up to undo the division above, approximating the true total loss (exact would have been a sum)
+        lossf = loss.item() * gradient_accumulation_steps
         if local_iter_num >= 5: # let the training loop settle a bit
             mfu = raw_model.estimate_mfu(batch_size * gradient_accumulation_steps, dt)
             running_mfu = mfu if running_mfu == -1.0 else 0.9*running_mfu + 0.1*mfu