support operator.index function #127440
Closed
support operator.index function #127440
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This was referenced May 29, 2024
mlazos
approved these changes
May 29, 2024
laithsakka
added
topic: not user facing
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
petrex
pushed a commit
to petrex/pytorch
that referenced
this pull request
Jun 5, 2024
Fix pytorch#127426 Pull Request resolved: pytorch#127440 Approved by: https://github.com/mlazos ghstack dependencies: pytorch#126444, pytorch#127146, pytorch#127424
pytorchmergebot
pushed a commit
that referenced
this pull request
Jun 6, 2024
number of call nodes increase due to inlining
before inlining:
```
class GraphModule(torch.nn.Module):
def forward(self, function_ctx, cat: "f32[1, s0, 512]"):
# No stacktrace found for following nodes
_set_grad_enabled = torch._C._set_grad_enabled(False)
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:283 in backward, code: grad_attn_output, grad_hidden_states = torch.chunk(
chunk = torch.chunk(cat, 2, dim = -1); cat = None
getitem: "f32[1, s0, 256]" = chunk[0]
getitem_1: "f32[1, s0, 256]" = chunk[1]; chunk = None
# No stacktrace found for following nodes
_set_grad_enabled_1 = torch._C._set_grad_enabled(True)
return (getitem_1, None)
```
after inlining:
```
class GraphModule(torch.nn.Module):
def forward(self, s0: "Sym(s0)", L_hidden_states_: "f32[1, s0, 256]", L_self_layers_0_weight: "f32[256, 256]", L_self_layers_0_bias: "f32[256]", L_self_layer_norm_weight: "f32[512]", L_self_layer_norm_bias: "f32[512]", L_self_layer_norm_normalized_shape_0_: "Sym(512)"):
l_hidden_states_ = L_hidden_states_
l_self_layers_0_weight = L_self_layers_0_weight
l_self_layers_0_bias = L_self_layers_0_bias
l_self_layer_norm_weight = L_self_layer_norm_weight
l_self_layer_norm_bias = L_self_layer_norm_bias
l_self_layer_norm_normalized_shape_0_ = L_self_layer_norm_normalized_shape_0_
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:332 in forward, code: hidden_states = torch.cat([hidden_states, hidden_states], dim=-1)
hidden_states: "f32[1, s0, 512]" = torch.cat([l_hidden_states_, l_hidden_states_], dim = -1); l_hidden_states_ = None
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:333 in forward, code: hidden_states = _ReversibleFunction.apply(
function_ctx = torch.autograd.function.FunctionCtx()
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:258 in forward, code: hidden_states, attn_output = torch.chunk(hidden_states, 2, dim=-1)
chunk = torch.chunk(hidden_states, 2, dim = -1); hidden_states = None
hidden_states_1: "f32[1, s0, 256]" = chunk[0]
attn_output: "f32[1, s0, 256]" = chunk[1]; chunk = None
# File: /data/users/lsakka/pytorch/pytorch/torch/nn/modules/linear.py:116 in forward, code: return F.linear(input, self.weight, self.bias)
attn_output_1: "f32[1, s0, 256]" = torch._C._nn.linear(attn_output, l_self_layers_0_weight, l_self_layers_0_bias); attn_output = l_self_layers_0_weight = l_self_layers_0_bias = None
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:272 in forward, code: ctx.save_for_backward(attn_output.detach(), hidden_states.detach())
detach: "f32[1, s0, 256]" = attn_output_1.detach()
detach_1: "f32[1, s0, 256]" = hidden_states_1.detach()
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:279 in forward, code: return torch.cat([attn_output, hidden_states], dim=-1)
hidden_states_2: "f32[1, s0, 512]" = torch.cat([attn_output_1, hidden_states_1], dim = -1); attn_output_1 = hidden_states_1 = None
# File: /data/users/lsakka/pytorch/pytorch/torch/nn/modules/normalization.py:201 in forward, code: return F.layer_norm(
hidden_states_3: "f32[1, s0, 512]" = torch.nn.functional.layer_norm(hidden_states_2, (l_self_layer_norm_normalized_shape_0_,), l_self_layer_norm_weight, l_self_layer_norm_bias, 1e-12); hidden_states_2 = l_self_layer_norm_normalized_shape_0_ = l_self_layer_norm_weight = l_self_layer_norm_bias = None
# File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:352 in forward, code: hidden_states = torch.nn.functional.dropout(
hidden_states_4: "f32[1, s0, 512]" = torch.nn.functional.dropout(hidden_states_3, p = 0.5, training = True); hidden_states_3 = None
return (hidden_states_4,)
```
Pull Request resolved: #127467
Approved by: https://github.com/anijain2305
ghstack dependencies: #126444, #127146, #127424, #127440
TharinduRusira
pushed a commit
to TharinduRusira/pytorch
that referenced
this pull request
Jun 14, 2024
…#127467) number of call nodes increase due to inlining before inlining: ``` class GraphModule(torch.nn.Module): def forward(self, function_ctx, cat: "f32[1, s0, 512]"): # No stacktrace found for following nodes _set_grad_enabled = torch._C._set_grad_enabled(False) # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:283 in backward, code: grad_attn_output, grad_hidden_states = torch.chunk( chunk = torch.chunk(cat, 2, dim = -1); cat = None getitem: "f32[1, s0, 256]" = chunk[0] getitem_1: "f32[1, s0, 256]" = chunk[1]; chunk = None # No stacktrace found for following nodes _set_grad_enabled_1 = torch._C._set_grad_enabled(True) return (getitem_1, None) ``` after inlining: ``` class GraphModule(torch.nn.Module): def forward(self, s0: "Sym(s0)", L_hidden_states_: "f32[1, s0, 256]", L_self_layers_0_weight: "f32[256, 256]", L_self_layers_0_bias: "f32[256]", L_self_layer_norm_weight: "f32[512]", L_self_layer_norm_bias: "f32[512]", L_self_layer_norm_normalized_shape_0_: "Sym(512)"): l_hidden_states_ = L_hidden_states_ l_self_layers_0_weight = L_self_layers_0_weight l_self_layers_0_bias = L_self_layers_0_bias l_self_layer_norm_weight = L_self_layer_norm_weight l_self_layer_norm_bias = L_self_layer_norm_bias l_self_layer_norm_normalized_shape_0_ = L_self_layer_norm_normalized_shape_0_ # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:332 in forward, code: hidden_states = torch.cat([hidden_states, hidden_states], dim=-1) hidden_states: "f32[1, s0, 512]" = torch.cat([l_hidden_states_, l_hidden_states_], dim = -1); l_hidden_states_ = None # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:333 in forward, code: hidden_states = _ReversibleFunction.apply( function_ctx = torch.autograd.function.FunctionCtx() # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:258 in forward, code: hidden_states, attn_output = torch.chunk(hidden_states, 2, dim=-1) chunk = torch.chunk(hidden_states, 2, dim = -1); hidden_states = None hidden_states_1: "f32[1, s0, 256]" = chunk[0] attn_output: "f32[1, s0, 256]" = chunk[1]; chunk = None # File: /data/users/lsakka/pytorch/pytorch/torch/nn/modules/linear.py:116 in forward, code: return F.linear(input, self.weight, self.bias) attn_output_1: "f32[1, s0, 256]" = torch._C._nn.linear(attn_output, l_self_layers_0_weight, l_self_layers_0_bias); attn_output = l_self_layers_0_weight = l_self_layers_0_bias = None # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:272 in forward, code: ctx.save_for_backward(attn_output.detach(), hidden_states.detach()) detach: "f32[1, s0, 256]" = attn_output_1.detach() detach_1: "f32[1, s0, 256]" = hidden_states_1.detach() # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:279 in forward, code: return torch.cat([attn_output, hidden_states], dim=-1) hidden_states_2: "f32[1, s0, 512]" = torch.cat([attn_output_1, hidden_states_1], dim = -1); attn_output_1 = hidden_states_1 = None # File: /data/users/lsakka/pytorch/pytorch/torch/nn/modules/normalization.py:201 in forward, code: return F.layer_norm( hidden_states_3: "f32[1, s0, 512]" = torch.nn.functional.layer_norm(hidden_states_2, (l_self_layer_norm_normalized_shape_0_,), l_self_layer_norm_weight, l_self_layer_norm_bias, 1e-12); hidden_states_2 = l_self_layer_norm_normalized_shape_0_ = l_self_layer_norm_weight = l_self_layer_norm_bias = None # File: /data/users/lsakka/pytorch/pytorch/test/dynamo/test_repros.py:352 in forward, code: hidden_states = torch.nn.functional.dropout( hidden_states_4: "f32[1, s0, 512]" = torch.nn.functional.dropout(hidden_states_3, p = 0.5, training = True); hidden_states_3 = None return (hidden_states_4,) ``` Pull Request resolved: pytorch#127467 Approved by: https://github.com/anijain2305 ghstack dependencies: pytorch#126444, pytorch#127146, pytorch#127424, pytorch#127440
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fix #127426
Stack from ghstack (oldest at bottom):
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang